# Appendix Source: https://www.recraft.ai/docs/api-reference/appendix ### Maximum prompt length The maximum prompt length depends on the model. | Models | Limit | | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | ----- | | Recraft V4.1
Recraft V4.1 Vector
Recraft V4.1 Pro
Recraft V4.1 Pro Vector
Recraft V4.1 Utility
Recraft V4.1 Utility Vector
Recraft V4.1 Utility Pro
Recraft V4.1 Utility Pro Vector
Recraft V4
Recraft V4 Vector
Recraft V4 Pro
Recraft V4 Pro Vector | 10000 | | Recraft V3
Recraft V3 Vector
Recraft V2
Recraft V2 Vector | 1000 | ### List of supported image sizes The image size can be specified using one of two formats: * Explicit dimensions (`WxH`): defines the exact width (`W`) and height (`H`) of the output image in pixels. Example: `1820x1024`. * Aspect ratio (`w:h`): defines the proportional relationship between width and height without specifying absolute dimensions. Example: `16:9`. See [Wikipedia](https://en.wikipedia.org/wiki/Aspect_ratio_\(image\)) for more details. The set of valid aspects and size values depends on the selected model. Refer to the table below for supported configurations.
Models Aspect Size
Recraft V4.1
Recraft V4.1 Utility
Recraft V4		 `1:1``1024x1024`
`2:1``1536x768`
`1:2``768x1536`
`3:2``1280x832`
`2:3``832x1280`
`4:3``1216x896`
`3:4``896x1216`
`5:4``1152x896`
`4:5``896x1152`
`6:10``832x1344`
`14:10``1280x896`
`10:14``896x1280`
`16:9``1344x768`
`9:16``768x1344`
Recraft V4.1 Pro
Recraft V4.1 Utility Pro
Recraft V4 Pro		 `1:1``2048x2048`
`2:1``3072x1536`
`1:2``1536x3072`
`3:2``2560x1664`
`2:3``1664x2560`
`4:3``2432x1792`
`3:4``1792x2432`
`5:4``2304x1792`
`4:5``1792x2304`
`6:10``1664x2688`
`14:10``2560x1792`
`10:14``1792x2560`
`16:9``2688x1536`
`9:16``1536x2688`
Recraft V2
Recraft V3		 `1:1``1024x1024`
`2:1``2048x1024`
`1:2``1024x2048`
`3:2``1536x1024`
`2:3``1024x1536`
`4:3``1365x1024`
`3:4``1024x1365`
`5:4``1280x1024`
`4:5``1024x1280`
`6:10``1024x1707`
`14:10``1434x1024`
`10:14``1024x1434`
`16:9``1820x1024`
`9:16``1024x1820`
Recraft V4.1 Vector
Recraft V4.1 Pro Vector
Recraft V4.1 Utility Vector
Recraft V4.1 Utility Pro Vector
Recraft V4 Vector
Recraft V4 Pro Vector
Recraft V3 Vector
Recraft V2 Vector		 `1:1`
`2:1`
`1:2`
`3:2`
`2:3`
`4:3`
`3:4`
`5:4`
`4:5`
`6:10`
`14:10`
`10:14`
`16:9`
`9:16`
### Policies * All generated images are currently stored for approx. 24 hours, this policy may change in the future, and you should not rely on it remaining constant. * Images are publicly accessible via direct links without authentication. However, since the URLs include unique image identifiers and are cryptographically signed, restoring lost links is nearly impossible. * Currently, image generation rates are defined on a per-user basis and set at **100 images per minute**. In addition, requests are limited to **5 per second**. These rate limits may be adjusted in the future. Need help or have suggestions for improving our docs? [Contact support](mailto:\[help@recraft.ai]\(mailto:help@recraft.ai\)) # Endpoints Source: https://www.recraft.ai/docs/api-reference/endpoints Dig into the details of the Recraft API endpoints. ## Authentication We use [Bearer](https://swagger.io/docs/specification/authentication/bearer-authentication/) API tokens for authentication. To access your API key, log in to Recraft and [enter your profile](https://app.recraft.ai/profile/api). All requests should include your API key in an Authorization HTTP header as follows:\ \ `Authorization: Bearer RECRAFT_API_TOKEN` ### OpenAI Python library Future examples will be shown using OpenAI Python library, for example, once installed, you can use the following code to be authenticated: ```python theme={null} from openai import OpenAI client = OpenAI( base_url='https://external.api.recraft.ai/v1', api_key=, ) ``` Since the Recraft API follows REST principles, you can interact with it using any standard HTTP client such as `curl`, as well as your preferred programming language or library. ## Image inputs: multipart or JSON Every endpoint that takes an image (or mask) as input accepts the request in **two interchangeable formats**: * **`multipart/form-data`** — upload the binary file directly using form fields (`image`, `mask`, `file`, ...). This is convenient for local files. * **`application/json`** — pass the image *by reference* instead of uploading bytes. Each file field has a JSON counterpart that accepts a **public URL** or a **[data URL](https://developer.mozilla.org/en-US/docs/Web/URI/Schemes/data)** (a `data:image/...;base64,...` string). The mapping between the multipart file fields and their JSON counterparts: | Multipart field | JSON field | Type | | ---------------------- | ------------ | -------------------------------------------------------- | | `image` | `image_url` | string — URL or data URL of the input image | | `mask` | `mask_url` | string — URL or data URL of the mask image | | `file` | `image_url` | string — URL or data URL of the input image | | `files` (Create style) | `image_urls` | array of strings — URLs or data URLs of the input images | All other parameters are identical between the two formats. Use JSON when your image is already hosted somewhere (just send the URL) or when a JSON body fits your client better; use multipart to upload local files directly. The examples below show both variants where applicable. ## Generate image Creates an image given a prompt. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/generations ``` In addition to the general endpoint above, two specialized variants accept the same request body but constrain the allowed `model` and `style` values to a single output type: ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/generations/raster POST https://external.api.recraft.ai/v1/images/generations/vector ``` * **`/generations/raster`** — only raster models (any model from the list above whose name does **not** end in `_vector`) and raster styles are accepted. Requests using a vector model or vector style are rejected. * **`/generations/vector`** — only vector models (any model from the list above whose name ends in `_vector`) and vector styles are accepted. Requests using a raster model or raster style are rejected. Use these variants when you want the server to enforce the output type — for example, in an agent or workflow that must produce only raster (or only vector) images. The base `/generations` endpoint remains available and accepts any supported model or style. ### Example ```python theme={null} response = client.images.generate( prompt='race car on a track', ) print(response.data[0].url) ``` ### Parameters | Parameter | Type | Description | Compatibility | | ----------------- | ---------------------------------------- | --------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | prompt (required) | string | A text description of the desired image(s). | Maximum prompt length depends on model, refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | All models | | model | string or null, default is `recraftv4_1` | The model to use for image generation. | Must be one of:
`recraftv4_1`
`recraftv4_1_vector`
`recraftv4_1_pro`
`recraftv4_1_pro_vector`
`recraftv4_1_utility`
`recraftv4_1_utility_vector`
`recraftv4_1_utility_pro`
`recraftv4_1_utility_pro_vector`
`recraftv4`
`recraftv4_vector`
`recraftv4_pro`
`recraftv4_pro_vector`
`recraftv3`
`recraftv3_vector`
`recraftv2`
`recraftv2_vector` | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | V2 / V3 styles | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | V2 / V3 styles | | size | string or null | The size of the generated images in `WxH` or `w:h` format. If not specified, the size is auto-selected based on the prompt. | Depends on model, supported values are published in [Appendix](/docs/api-reference/appendix#list-of-supported-image-sizes) | | negative\_prompt | string or null | A text description of undesired elements on an image. | V2 / V3 models | | random\_seed | integer or null | Seed for reproducible generation. | All models | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | All models | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | V3 models only | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | All models, partially supported for V4 | **Hint:** if OpenAI Python Library is used, non-standard parameters can be passed using the `extra_body` argument. For example: ```python theme={null} response = client.images.generate( prompt='race car on a track', extra_body={ 'style_id': style_id, 'controls': { ... } } ) print(response.data[0].url) ``` ## Create style Upload a set of images to create a style reference. ```javascript theme={null} POST https://external.api.recraft.ai/v1/styles ``` ### Example ```python Multipart theme={null} response = client.post( path='/styles', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'style': 'digital_illustration'}, files={'file1': open('image.png', 'rb')}, ) print(response['id']) ``` ```python JSON theme={null} response = client.post( path='/styles', cast_to=object, body={ 'style': 'digital_illustration', 'image_urls': ['https://example.com/image.png'], }, ) print(response['id']) ``` ### Output ```javascript theme={null} { "id": "229b2a75-05e4-4580-85f9-b47ee521a00d", "style": "digital_illustration", "creation_time": "2024-01-01T00:00:00Z", "is_private": true, "credits": 40 } ``` ### Request body Upload a set of images to create a style reference. | Parameter | Type | Description | | ---------------------- | ------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | style (required) | string | The base style of the generated images, should be one of: `any`, `realistic_image`, `digital_illustration`, `vector_illustration`, `icon`, `logo_raster` | | files | files | **(multipart only)** Images in PNG, JPG, or WEBP format for use as style references, sent as multipart file parts (any field name). The max number of images is 5. Total size of all images is limited to 5 MB (each file must also be under the 10 MB upload limit). Required unless `source_styles` is provided. | | image\_urls | array of strings | **(JSON only)** URLs or data URLs of the reference images, the JSON-body counterpart of `files`. Same image format, count, and size limits apply. Required unless `source_styles` is provided. | | model | string or null | The transform model the style is created for, e.g. `recraftv3`, `recraftv3_vector`. | | image\_weights | array of numbers or null | Per-image weights; the number of weights must match the number of uploaded images. | | source\_styles | array of UUIDs or null | Existing style IDs to mix into the new style. May be used instead of, or together with, uploaded images. | | source\_style\_weights | array of numbers or null | Per-source-style weights; the number of weights must match the number of `source_styles`. | | prompt | string or null | Optional text prompt associated with the style. | | mix\_policy | string or null | How sources are mixed when combining images/source styles. Must be one of `PaletteMatch`, `MaxWeight`. | ## Image to image Image-to-image operation tool allows you to create images similar to a given image, preserving certain aspects like composition, color, or subject identity while altering others based on the prompt. This can be used to create variations of the image or images that have similar composition to existing image. Use prompt to describe the new image and strength to define similarity level. ***Note***: This operation is available with Recraft V3 and Recraft V4 models (raster and vector). ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/imageToImage ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/imageToImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', 'strength': 0.2, }, ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/imageToImage', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'prompt': 'winter', 'strength': 0.2, }, ) print(response['data'][0]['url']) ``` ### Parameters In JSON mode, replace the `image` file with `image_url` (a URL or data URL of the input image); all other parameters are the same. | Parameter | Type | Description | Compatibility | | --------------------- | ---------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | image (required) | file | **(multipart)** An image to modify, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. PNG, JPG, WEBP, or SVG. The output format is determined by the selected model. | | | image\_url (required) | string | **(JSON)** URL or data URL of the input image, the JSON-body counterpart of `image`. Same format and size limits apply. | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | strength (required) | float | Defines the difference with the original image, should lie in `[0, 1]`, where `0` means almost identical, and `1` means minimal similarity. | | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv4_1` | The model to use for image generation. | Must be one of the Recraft V3 / V4 models: `recraftv3`, `recraftv3_vector`, `recraftv4`, `recraftv4_vector`, `recraftv4_pro`, `recraftv4_pro_vector`, `recraftv4_1`, `recraftv4_1_vector`, `recraftv4_1_pro`, `recraftv4_1_pro_vector`, `recraftv4_1_utility`, `recraftv4_1_utility_vector`, `recraftv4_1_utility_pro`, `recraftv4_1_utility_pro_vector` | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | V3 / V4 styles | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | V3 / V4 styles | | random\_seed | integer or null | Seed for reproducible generation. | | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Image inpainting Inpainting lets you regenerate specific parts of an image while keeping the rest intact. This is useful for correcting details, replacing elements, or making creative changes without starting from scratch. It uses a mask to identify the areas to be filled in, where white pixels represent the regions to inpaint, and black pixels indicate the areas to keep intact, i.e. the white pixels are filled based on the input provided in the prompt. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/inpaint ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/inpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/inpaint', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'mask_url': 'https://example.com/mask.png', 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters The request can be sent either as `multipart/form-data` (uploading the `image` and `mask` files) or as `application/json` (passing `image_url` and `mask_url` instead). The image must be in PNG, JPG, WEBP, or SVG format and the mask in PNG, JPG, or WEBP. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. The output format (raster or vector) is determined by the selected model. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | mask (required) | file | An image encoded in **grayscale color mode**, used to define the specific regions of an image that need modification. The white pixels represent the parts of the image that will be inpainted, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Image outpainting Outpainting extends an image beyond its original boundaries by generating new content around it. This is useful for changing the aspect ratio, adding scenery, or zooming out to reveal more of the scene while keeping the original image intact. The area to be generated can be specified either by per-side pixel margins via `expand_left`, `expand_right`, `expand_top`, `expand_bottom` (which extend the image by the given amount on each side), or by a target `size` (which places the source image inside the resulting image of the given dimensions); these two options cannot be used together. After that, it is also possible to specify `zoom_out_percentage`, which scales the source image down relative to the resulting image by the given percentage and can be combined with either option or used on its own. At least one of `size`, `expand_*`, or `zoom_out_percentage` must be specified. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/outpaint ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/outpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'a mountain landscape', 'size': '16:9', 'zoom_out_percentage': 25.0, }, ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/outpaint', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'prompt': 'a mountain landscape', 'size': '16:9', 'zoom_out_percentage': 25.0, }, ) print(response['data'][0]['url']) ``` ### Parameters The request can be sent either as `multipart/form-data` (uploading the `image` file) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, WEBP, or SVG format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. The resulting image dimensions (after applying `expand_*`) must not exceed 4096 pixels on either side. The output format (raster or vector) is determined by the selected model. | Parameter | Type | Description | Compatibility | | --------------------- | -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------------------------- | | image (required) | file | An image to extend, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | prompt (required) | string | A text description of the content to generate in the extended area. | refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | expand\_left | integer or null | Number of pixels to add to the left side of the image, must be in range `[0, 4096]`. Cannot be combined with `size`. | | | expand\_right | integer or null | Number of pixels to add to the right side of the image, must be in range `[0, 4096]`. Cannot be combined with `size`. | | | expand\_top | integer or null | Number of pixels to add to the top side of the image, must be in range `[0, 4096]`. Cannot be combined with `size`. | | | expand\_bottom | integer or null | Number of pixels to add to the bottom side of the image, must be in range `[0, 4096]`. Cannot be combined with `size`. | | | size | string or null | The size of the resulting outpainted image in `WxH` or `w:h` format. The source image is placed inside the resulting image. Cannot be combined with `expand_*`. | Depends on model, supported values are published in [Appendix](/docs/api-reference/appendix#list-of-supported-image-sizes) | | zoom\_out\_percentage | float or null, default is `0` | Percentage by which the source image is scaled down before outpainting, must be in range `[0, 100)`. Higher values produce more surrounding context. Can be used on its own or combined with `size` or `expand_*`. | | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the content generated in the extended area, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference for the content generated in the extended area, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Replace background Replace Background operation detects background of an image and modifies it according to given prompt. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/replaceBackground ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/replaceBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/replaceBackground', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters The request can be sent either as `multipart/form-data` (uploading the `image` file) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, WEBP, or SVG format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. The output format (raster or vector) is determined by the selected model. | Parameter | Type | Description | Compatibility | | --------------------- | -------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | **(multipart)** An image to modify, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | image\_url (required) | string | **(JSON)** URL or data URL of the input image, the JSON-body counterpart of `image`. | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Generate background Generate Background operation generates a background for a given image, based on a prompt and a mask that specifies the regions to fill. ***Note***: This operation is available with Recraft V3, Recraft V3 Vector models only. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/generateBackground ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/generateBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/generateBackground', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'mask_url': 'https://example.com/mask.png', 'prompt': 'winter', }, ) print(response['data'][0]['url']) ``` ### Parameters The request can be sent either as `multipart/form-data` (uploading the `image` and `mask` files) or as `application/json` (passing `image_url` and `mask_url` instead). The image must be in PNG, JPG, WEBP, or SVG format and the mask in PNG, JPG, or WEBP. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. The output format (raster or vector) is determined by the selected model. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------ | | image (required) | file | An image to modify, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | | mask (required) | file | An image encoded in **grayscale color mode**, used to define the specific regions of an image that need modification. The white pixels represent the parts of the image that will be inpainted, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | | prompt (required) | string | A text description of areas to change. | refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length) | | n | integer or null, default is 1 | The number of images to generate, must be between 1 and 6. | | | model | string or null, default is `recraftv3` | The model to use for image generation. | Must be one of: `recraftv3`, `recraftv3_vector` | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | V3 styles only | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | | negative\_prompt | string or null | A text description of undesired elements on an image. | | | text\_layout | Array of objects or null | Refer to [Text Layout](#text-layout). | | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). | | ## Vectorize image Converts a given raster image to SVG format. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/vectorize ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/vectorize', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ```python JSON theme={null} response = client.post( path='/images/vectorize', cast_to=object, body={'image_url': 'https://example.com/image.png'}, ) print(response['image']['url']) ``` ### **Parameters** The request can be sent either as `multipart/form-data` (uploading the image as the `file` field) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, or WEBP format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Remove background Removes background of a given image. Accepts raster (PNG, JPG, WEBP) or SVG input. When the input is an SVG, the result is returned as an SVG (vector output); raster input returns a raster image. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/removeBackground ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/removeBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ```python JSON theme={null} response = client.post( path='/images/removeBackground', cast_to=object, body={'image_url': 'https://example.com/image.png'}, ) print(response['image']['url']) ``` ### **Parameters** The request can be sent either as `multipart/form-data` (uploading the image as the `file` field) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, WEBP, or SVG format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. SVG inputs smaller than the 256 px minimum are scaled up before processing; SVGs exceeding the max dimension/resolution, and unsafe SVGs, are rejected. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Crisp upscale Enhances a given raster image using ‘crisp upscale’ tool, increasing image resolution, making the image sharper and cleaner. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/crispUpscale ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/crispUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ```python JSON theme={null} response = client.post( path='/images/crispUpscale', cast_to=object, body={'image_url': 'https://example.com/image.png'}, ) print(response['image']['url']) ``` ### Request body The request can be sent either as `multipart/form-data` (uploading the image as the `file` field) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, or WEBP format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 32 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Creative upscale Enhances a given raster image using ‘creative upscale’ tool, boosting resolution with a focus on refining small details and faces. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/creativeUpscale ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/creativeUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ```python JSON theme={null} response = client.post( path='/images/creativeUpscale', cast_to=object, body={'image_url': 'https://example.com/image.png'}, ) print(response['image']['url']) ``` ### Request body The request can be sent either as `multipart/form-data` (uploading the image as the `file` field) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, or WEBP format. The image must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. | Parameter | Type | Description | | ---------------- | -------------------------------- | ------------------------------------------------------------------------------------------ | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Erase region Erases a region of a given raster image following a given mask, where white pixels represent the regions to erase, and black pixels indicate the areas to keep intact. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/eraseRegion ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/eraseRegion', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb')}, ) print(response['image']['url']) ``` ```python JSON theme={null} response = client.post( path='/images/eraseRegion', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'mask_url': 'https://example.com/mask.png', }, ) print(response['image']['url']) ``` ### Request body The request can be sent either as `multipart/form-data` (uploading the `image` and `mask` files) or as `application/json` (passing `image_url` and `mask_url` instead). The image must be in PNG, JPG, WEBP, or SVG format and the mask in PNG, JPG, or WEBP. The images must be no more than 10 MB in size, have resolution no more than 16 MP, max dimension no more than 4096 pixels and min dimension no less than 256 pixels. The mask and image must have the same dimensions. When the input is an SVG, the result is returned as an SVG; a raster input returns a raster image. | Parameter | Type | Description | | --------------------- | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | image (required) | file | **(multipart)** An image to modify, must be less than 10 MB in size, have resolution less than 16 MP, and max dimension less than 4096 pixels. | | image\_url (required) | string | **(JSON)** URL or data URL of the input image, the JSON-body counterpart of `image`. | | mask (required) | file | **(multipart)** An image encoded in **grayscale color mode**, used to define the specific regions of the image to be erased. The white pixels represent the parts of the image that will be erased, while black pixels indicate the parts of the image that will remain unchanged. Should have exactly the same size as the image. Each pixel of the image should be either pure black (value `0`) or pure white (value `255`). | | mask\_url (required) | string | **(JSON)** URL or data URL of the mask image, the JSON-body counterpart of `mask`. | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Remix image Remix generates new versions of an image while keeping the overall concept intact. Each remix introduces slight differences in elements like character appearance, object position, or scene composition. The Remix function does not use a prompt. Instead, it uses the visual content of the original image to generate alternatives in different aspect ratios. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/variateImage ``` ### Example ```python Multipart theme={null} response = client.post( path='/images/variateImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'image': open('image.png', 'rb')}, body={'size': '1024x1024'} ) print(response['data'][0]['url']) ``` ```python JSON theme={null} response = client.post( path='/images/variateImage', cast_to=object, body={ 'image_url': 'https://example.com/image.png', 'size': '1024x1024', }, ) print(response['data'][0]['url']) ``` ### Parameters The request can be sent either as `multipart/form-data` (uploading the `image` file) or as `application/json` (passing `image_url` — a URL or data URL — instead). The input image must be in PNG, JPG, WEBP, or SVG format. The image must not exceed 10 MB in size, with a maximum resolution of 16 MP, a maximum dimension of 4096 px, and a minimum dimension of 256 px. The output format (raster or vector) is determined by the selected model. | Parameter | Type | Description | | --------------------- | -------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------- | | image (required) | file | **(multipart)** The input image in PNG, WEBP, JPEG, or SVG format. | | image\_url (required) | string | **(JSON)** URL or data URL of the input image, the JSON-body counterpart of `image`. | | size (required) | string | The size of the generated images in `WxH` or `w:h` format, supported values are published in [Appendix](/docs/api-reference/appendix#list-of-supported-image-sizes). | | n | integer or null, default is 1 | Number of variations to generate \[1–6]. | | model | string or null | The model to use for generation. The output format (raster or vector) is determined by the selected model. | | random\_seed | integer or null | Optional random seed for reproducibility. | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Explore Generates a set of images based on a prompt, designed for visual exploration and discovery of diverse results. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/explore ``` ### Example ```python theme={null} response = client.post( path='/images/explore', cast_to=object, body={ 'prompt': 'race car on a track', }, ) print(response['data'][0]['url']) ``` ### Parameters | Parameter | Type | Description | | ----------------- | ---------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | prompt (required) | string | A text description of the desired image(s). Maximum prompt length depends on model, refer to [Appendix](/docs/api-reference/appendix#maximum-prompt-length). | | model | string or null, default is `recraftv4_1` | The model to use for image generation. Must be one of: `recraftv4_1`, `recraftv4_1_vector`, `recraftv4_1_pro`, `recraftv4_1_pro_vector`, `recraftv4`, `recraftv4_vector`, `recraftv4_pro`, `recraftv4_pro_vector`. | | style | string or null | The style of the generated images, refer to [Styles](/docs/api-reference/styles). | | style\_id | UUID or null | Use a style as visual reference, refer to [Styles](/docs/api-reference/styles). | | size | string or null, default is `1:1` | The size of the generated images in `WxH` or `w:h` format. Depends on model, supported values are published in [Appendix](/docs/api-reference/appendix#list-of-supported-image-sizes). | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | | controls | object or null | A set of custom parameters to tweak generation process, refer to [Controls](#controls). Supports `prompt_to_image` controls. | ## Explore similar Start from an image you already created and generate new images that are visually similar to the given source image. The source image **must** be one previously generated via the Explore endpoint. The `similarity` parameter controls how closely the results should resemble the source. ```javascript theme={null} POST https://external.api.recraft.ai/v1/images/explore/similar ``` ### Example ```python theme={null} response = client.post( path='/images/explore/similar', cast_to=object, body={ 'source_image_id': 'c18a1988-45e7-4c00-82c4-4ad7d3dbce3a', 'similarity': 5, }, ) print(response['data'][0]['url']) ``` ### Parameters | Parameter | Type | Description | | ---------------------------- | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | source\_image\_id (required) | UUID | The ID of the source image to use as a visual reference for exploration. | | similarity (required) | integer | Defines how closely the generated images should resemble the source image. The value should be between `1` and `5`: `1` — a little bit similar, `2` — moderately similar, `3` — quite similar, `4` — very similar, `5` — extremely similar. | | response\_format | string or null, default is `url` | The format in which the generated images are returned. Must be one of `url` or `b64_json`. | ## Enhance prompt Expands a short prompt into a richer, more detailed description with added visual context, style cues, and composition details. Use the enhanced prompt in any image generation endpoint to produce more vivid and accurate results. ```javascript theme={null} POST https://external.api.recraft.ai/v1/prompts/enhance ``` ### Example ```python theme={null} response = client.post( path='/prompts/enhance', cast_to=object, body={'prompt': 'race car on a track'}, ) print(response['enhanced_prompt']) ``` ### Parameters | Parameter | Type | Description | | ----------------- | ------ | ---------------------------------------------------------------------------------------- | | prompt (required) | string | The original prompt to enhance. Must not be empty and must not exceed `2000` characters. | ## Get user information Returns information of the current user including credits balance. ```javascript theme={null} GET https://external.api.recraft.ai/v1/users/me ``` ### Example ```python theme={null} response = client.get(path='/users/me', cast_to=object) print(response) ``` ### Output ```javascript theme={null} { "credits": 1000, "email": "test@example.com", "id": "c18a1988-45e7-4c00-82c4-4ad7d3dbce3a", "name": "Recraft Test" } ``` ## Auxiliary ### Controls The generation process can be adjusted with a number of tweaks. | Parameter | Type | Description | Compatibility | | ----------------- | -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | -------------- | | colors | array of color definitions | An array of preferable colors. | All models | | background\_color | color definition | Use the given color as a desired background color. | All models | | artistic\_level | integer or null | Defines the artistic tone of your image. At a simple level, the person looks straight at the camera in a static and clean style. Dynamic and eccentric levels introduce movement and creativity. The value should be in the range `[0..5]`. | V3 models only | | no\_text | bool | Do not embed text layouts. | V3 models only | #### Colors Color type is defined as an object with the following fields | Parameter | Description | | -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- | | rgb (required) | An array of 3 integer values in range of `0...255` defining RGB color model. | | weight | An optional number in range `0.0...1.0` indicating the relative preference of this color. The sum of weights across all `colors` must not exceed `1.0`. | **Example** ```python theme={null} response = client.images.generate( prompt='race car on a track', extra_body={ 'controls': { 'colors': [ {'rgb': [0, 255, 0]} ] } } ) print(response.data[0].url) ``` ### Text Layout Text layout is used to define spatial and textual attributes for individual text elements. Each text element consists of an individual word and its bounding box (bbox). This feature is supported by Recraft V3 and Recraft V3 Vector models only. | Parameter | Description | | --------------- | ------------------------------------------------------------------------------------------------------------- | | text (required) | A single word containing only supported characters. | | bbox (required) | A bounding box representing a 4-angled polygon. Each point in the polygon is defined by relative coordinates. | **Bounding box**: The bounding box (bbox) is a list of 4 points representing a 4-angled figure (not necessarily a rectangle). Each point specifies its coordinates relative to the layout dimensions, where (0, 0) is the top-left corner, (1, 1) is the bottom-right corner. **Coordinates**: Coordinates are **relative** to the layout dimensions. Coordinates can extend beyond the \[0, 1] range, such values indicate that the shape will be cropped. **Points**: The bounding box must always have **exactly 4 points**. Each point is an array of two floats \[x, y] representing the relative position. **Supported characters** The `text` field must contain a single word composed only of the following characters: ```text theme={null} ! " # $ % & ' ( ) * + , - . / 0 1 2 3 4 5 6 7 8 9 : ; < > ? @ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z _ { } Ø Đ Ħ Ł Ŋ Ŧ Α Β Ε Ζ Η Ι Κ Μ Ν Ο Ρ Τ Υ Χ І А В Е К М Н О Р С Т У Х ß ẞ ``` Any character not listed above will result in validation errors. **Example** ```python theme={null} response = client.images.generate( prompt="cute red panda with a sign", style="Illustration", extra_body={ "text_layout": [ { "text": "Recraft", "bbox": [[0.3, 0.45], [0.6, 0.45], [0.6, 0.55], [0.3, 0.55]], }, { "text": "AI", "bbox": [[0.62, 0.45], [0.70, 0.45], [0.70, 0.55], [0.62, 0.55]], }, ] }, ) print(response.data[0].url) ``` # Examples Source: https://www.recraft.ai/docs/api-reference/examples Generate AI images using cURL or Python and create your own styles programmatically. ### Generate a raster image using Recraft V4.1 model ```bash generate_recraftv4_1.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "two race cars on a track", "model": "recraftv4_1" }' ``` ```python generate_recraftv4_1.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='two race cars on a track', model='recraftv4_1' ) print(response.data[0].url) ``` ### Generate a vector image using Recraft V4.1 Vector model ```bash generate_recraftv4_1_vector.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "cat on a mat", "model": "recraftv4_1_vector" }' ``` ```python generate_recraftv4_1_vector.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='cat on a mat', model='recraftv4_1_vector' ) print(response.data[0].url) ``` ### Generate an image using Recraft V4.1 Pro with specific size ```bash generate_recraftv4_1_pro_with_size.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "red point siamese cat", "model": "recraftv4_1_pro", "size": "2560x1664" }' ``` ```python generate_recraftv4_1_pro_with_size.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='red point siamese cat', model='recraftv4_1_pro', size='2560x1664', ) print(response.data[0].url) ``` ### Generate a digital illustration by Recraft V3 with specific style ```bash generate_digital_illustration.sh theme={null} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "a monster with lots of hands", "style": "Hand-drawn" }' ``` ```python generate_digital_illustration.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.images.generate( prompt='a monster with lots of hands', style='Hand-drawn', ) print(response.data[0].url) ``` ### Image to image by Recraft V3 with digital illustration style ```bash image_to_image.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/imageToImage \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "image=@image.png" \ -F "prompt=winter" \ -F "strength=0.2" \ -F "style=Illustration" ``` ```bash image_to_image_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/imageToImage \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "image_url": "https://example.com/image.png", "prompt": "winter", "strength": 0.2, "style": "Illustration" }' ``` ```python image_to_image.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/imageToImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'winter', 'strength': 0.2, 'style': 'Illustration', }, ) print(response['data'][0]['url']) ``` ### Inpaint an image by Recraft V3 with style Enterprise ```bash inpaint.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/inpaint \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "prompt=moon" \ -F "style=Enterprise" \ -F "image=@image.png" -F "mask=@mask.png" ``` ```bash inpaint_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/inpaint \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "moon", "style": "Enterprise", "image_url": "https://example.com/image.png", "mask_url": "https://example.com/mask.png" }' ``` ```python inpaint.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/inpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), 'mask': open('mask.png', 'rb'), }, body={ 'style': 'Enterprise', 'prompt': 'moon', }, ) print(response['data'][0]['url']) ``` ### Outpaint an image by Recraft V3 to a 16:9 aspect ratio ```bash outpaint.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/outpaint \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "prompt=a mountain landscape" \ -F "size=16:9" \ -F "zoom_out_percentage=25.0" \ -F "image=@image.png" ``` ```bash outpaint_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/outpaint \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "a mountain landscape", "size": "16:9", "zoom_out_percentage": 25.0, "image_url": "https://example.com/image.png" }' ``` ```python outpaint.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/outpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'a mountain landscape', 'size': '16:9', 'zoom_out_percentage': 25.0, }, ) print(response['data'][0]['url']) ``` ### Outpaint an image by Recraft V3 with per-side pixel margins ```bash outpaint_expand.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/outpaint \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "prompt=a mountain landscape" \ -F "expand_left=256" \ -F "expand_right=256" \ -F "expand_top=128" \ -F "expand_bottom=128" \ -F "image=@image.png" ``` ```bash outpaint_expand_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/outpaint \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "a mountain landscape", "expand_left": 256, "expand_right": 256, "expand_top": 128, "expand_bottom": 128, "image_url": "https://example.com/image.png" }' ``` ```python outpaint_expand.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/outpaint', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={ 'image': open('image.png', 'rb'), }, body={ 'prompt': 'a mountain landscape', 'expand_left': 256, 'expand_right': 256, 'expand_top': 128, 'expand_bottom': 128, }, ) print(response['data'][0]['url']) ``` ### Create own Recraft V3 style by uploading reference images and use them for generation ```bash create_style_and_generate.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/styles \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "style=digital_illustration" \ -F "file=@image.png" # response: {"id":"095b9f9d-f06f-4b4e-9bb2-d4f823203427"} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "wood potato masher", "style_id": "095b9f9d-f06f-4b4e-9bb2-d4f823203427" }' ``` ```bash create_style_and_generate_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/styles \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "style": "digital_illustration", "image_urls": ["https://example.com/image.png"] }' # response: {"id":"095b9f9d-f06f-4b4e-9bb2-d4f823203427"} curl https://external.api.recraft.ai/v1/images/generations \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "wood potato masher", "style_id": "095b9f9d-f06f-4b4e-9bb2-d4f823203427" }' ``` ```python create_style_and_generate.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) style = client.post( path='/styles', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'style': 'digital_illustration'}, files={'file': open('image.png', 'rb')}, ) print(style['id']) response = client.images.generate( prompt='wood potato masher', extra_body={'style_id': style['id']}, ) print(response.data[0].url) ``` ### Vectorize an image in PNG format ```bash vectorize.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/vectorize \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "file=@image.png" ``` ```bash vectorize_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/vectorize \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "image_url": "https://example.com/image.png" }' ``` ```python vectorize.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/vectorize', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Remove background from a PNG image, get the result in B64 JSON ```bash remove_background_b64.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/removeBackground \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=b64_json" \ -F "file=@image.png" ``` ```bash remove_background_b64_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/removeBackground \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "response_format": "b64_json", "image_url": "https://example.com/image.png" }' ``` ```python remove_background_b64.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/removeBackground', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'response_format': 'b64_json'}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Run crisp upscale tool for a PNG image, get the result in B64 JSON ```bash crisp_upscale_b64.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/crispUpscale \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=b64_json" \ -F "file=@image.png" ``` ```bash crisp_upscale_b64_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/crispUpscale \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "response_format": "b64_json", "image_url": "https://example.com/image.png" }' ``` ```python crisp_upscale_b64.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/crispUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={'response_format': 'b64_json'}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Run creative upscale tool for a PNG image ```bash creative_upscale.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/creativeUpscale \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "file=@image.png" ``` ```bash creative_upscale_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/creativeUpscale \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "image_url": "https://example.com/image.png" }' ``` ```python creative_upscale.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/creativeUpscale', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, files={'file': open('image.png', 'rb')}, ) print(response['image']['url']) ``` ### Variate PNG image, get the result in WEBP format ```bash variate_image.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/variateImage \ -H "Content-Type: multipart/form-data" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -F "response_format=url" \ -F "size=1024x1024" \ -F "n=1" \ -F "seed=13191922" \ -F "image_format=webp" \ -F "file=@image.png" ``` ```bash variate_image_json.sh theme={null} curl -X POST https://external.api.recraft.ai/v1/images/variateImage \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "response_format": "url", "size": "1024x1024", "n": 1, "seed": 13191922, "image_format": "webp", "image_url": "https://example.com/image.png" }' ``` ```python variate_image.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/variateImage', cast_to=object, options={'headers': {'Content-Type': 'multipart/form-data'}}, body={"size": "1024x1024", "n": 1, "response_format": "url", "seed": 13191922, "image_format": "webp"}, files={'file': open('image.png', 'rb')}, ) print(response['data'][0]["url"]) ``` ### Explore images ```bash explore.sh theme={null} curl https://external.api.recraft.ai/v1/images/explore \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "race car on a track", "model": "recraftv4" }' ``` ```python explore.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/explore', cast_to=object, body={ 'prompt': 'race car on a track', 'model': 'recraftv4', }, ) print(response['data'][0]['url']) ``` ### Enhance a prompt ```bash enhance_prompt.sh theme={null} curl https://external.api.recraft.ai/v1/prompts/enhance \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "prompt": "race car on a track" }' ``` ```python enhance_prompt.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/prompts/enhance', cast_to=object, body={'prompt': 'race car on a track'}, ) print(response['enhanced_prompt']) ``` ### Explore similar images ```bash explore_similar.sh theme={null} curl https://external.api.recraft.ai/v1/images/explore/similar \ -H "Content-Type: application/json" \ -H "Authorization: Bearer $RECRAFT_API_TOKEN" \ -d '{ "source_image_id": "c18a1988-45e7-4c00-82c4-4ad7d3dbce3a", "similarity": 3 }' ``` ```python explore_similar.py theme={null} from openai import OpenAI client = OpenAI(base_url='https://external.api.recraft.ai/v1', api_key=_RECRAFT_API_TOKEN) response = client.post( path='/images/explore/similar', cast_to=object, body={ 'source_image_id': 'c18a1988-45e7-4c00-82c4-4ad7d3dbce3a', 'similarity': 3, }, ) print(response['data'][0]['url']) ``` # Getting started Source: https://www.recraft.ai/docs/api-reference/getting-started Welcome to Recraft image generation and editing API. Learn the basics of the Recraft API, including raster and vector image generation, style creation, image generation in your brand style and colors, image vectorization, and background removal. **Authenticate and interact with our API in a matter of minutes.** ### Models Recraft has developed four generations of proprietary models: | Generation | Models | | ---------- | ------------------------------------------------------------------------------------------------------------------------------------------------ | | V4.1 | Recraft V4.1
Recraft V4.1 Vector
Recraft V4.1 Pro
Recraft V4.1 Pro Vector
Recraft V4.1 Utility
Recraft V4.1 Utility Pro | | V4 | Recraft V4
Recraft V4 Vector
Recraft V4 Pro
Recraft V4 Pro Vector | | V3 | Recraft V3 (aka Red Panda)
Recraft V3 Vector | | V2 | Recraft V2 (aka Recraft 20B)
Recraft V2 Vector | Recraft V4.1 models, released in mid-May 2026, are the next step beyond V4, with sharper, more polished, and more consistent results across the board. Photorealism takes a serious step forward — cleaner outputs, better object understanding, and noticeably sharper mockups. Prompt following has improved too: V4.1 reads intent more naturally, so shorter prompts produce stronger results. New illustration styles are now possible, and icons and vectors come out cleaner by default. V4.1 ships in two lines — V4.1, tuned for peak artistic and design quality, and V4.1 Utility, optimized for broader, general-purpose appeal. Each line is available in standard (1MP) versions for everyday work and fast iteration, and Pro (4MP) versions for print-ready assets and large-scale use. Recraft V4 models, released in February 2026, represent a ground-up rebuild and our most advanced generation to date. Focused on what matters for real work — design taste, prompt accuracy, and output quality at any size — V4 models deliver true design vision in composition, lighting, textures, and overall feel, along with the world's only production-grade vector generation. Available in two versions: V4 (1MP) for everyday work and fast iteration, and V4 Pro (4MP) for print-ready assets and large-scale use. Both share the same creative capabilities, design taste, and art-directed professional results. Recraft V3 models, released in October 2024, were new state-of-the-art models trained from scratch that introduced major advances in photorealism and text rendering. V3 ranked first on the Hugging Face Text-to-Image leaderboard for five consecutive months. Recraft V2 models, released in February 2024, were the first AI models built specifically for designers. V2 enabled the creation of both raster and vector images with anatomical accuracy, consistent brand styling, and precise iteration control. All 16 models from V2, V3, V4, V4.1 generations are available in the Recraft API and within the Recraft web and mobile applications. ## Supported features The Recraft API exposes the following capabilities. See [Endpoints](/docs/api-reference/endpoints) for full request and response schemas. * [Generate image](/docs/api-reference/endpoints#generate-image) — create raster or vector images from a text prompt. * [Create style](/docs/api-reference/endpoints#create-style) — upload reference images to define a reusable custom style. * [Image to image](/docs/api-reference/endpoints#image-to-image) — generate variations of an existing image guided by a prompt. * [Image inpainting](/docs/api-reference/endpoints#image-inpainting) — regenerate masked regions of an image while keeping the rest intact. * [Image outpainting](/docs/api-reference/endpoints#image-outpainting) — extend an image beyond its original borders. * [Replace background](/docs/api-reference/endpoints#replace-background) — swap the background of an image with a new one described by a prompt. * [Generate background](/docs/api-reference/endpoints#generate-background) — generate a fresh background around a subject. * [Vectorize image](/docs/api-reference/endpoints#vectorize-image) — convert a raster image into a scalable vector graphic. * [Remove background](/docs/api-reference/endpoints#remove-background) — produce a transparent-background cutout of the subject. * [Crisp upscale](/docs/api-reference/endpoints#crisp-upscale) — increase resolution without altering image content. * [Creative upscale](/docs/api-reference/endpoints#creative-upscale) — increase resolution while regenerating finer details. * [Erase region](/docs/api-reference/endpoints#erase-region) — remove selected objects or regions from an image. * [Remix image](/docs/api-reference/endpoints#remix-image) — recompose an image with adjustable variation strength. * [Explore](/docs/api-reference/endpoints#explore) — generate a grid of prompt variations to explore visual directions. * [Explore similar](/docs/api-reference/endpoints#explore-similar) — generate further variations close to a chosen reference image. * [Enhance prompt](/docs/api-reference/endpoints#enhance-prompt) — rewrite a short prompt into a richer, more detailed one. * [Get user information](/docs/api-reference/endpoints#get-user-information) — fetch account details and remaining balance. ## Authentication We use [Bearer](https://swagger.io/docs/specification/authentication/bearer-authentication/) API tokens for authentication. To access your API key, log in to Recraft, [enter your profile](https://app.recraft.ai/profile/api) and hit 'Generate' (available only if your API units balance is above zero). All requests should include your API key in an Authorization HTTP header as follows:\ \ `Authorization: Bearer RECRAFT_API_TOKEN` A user may create multiple API tokens; however, all tokens share the same API units balance. ## Swagger Interactive API documentation is available via [Swagger](https://external.api.recraft.ai/doc/#/). You can explore all available endpoints, test API calls, and view request/response schemas directly in your browser. ### REST / Python Library The Recraft API adheres to REST principles, allowing you to interact using any utilities (e.g., curl), programming languages, or libraries of your choice. One of the easiest of available alternatives is [OpenAI Python library](https://github.com/openai/openai-python) which is also compatible with Recraft API, but it’s important to remember that not all parameters/options are supported or implemented. Additionally, some parameters may have different meanings, or they may be quietly ignored if they are not applicable to the Recraft API. Future examples will be shown using that library, for example, once installed, you can use the following code to be authenticated: ```python theme={null} from openai import OpenAI client = OpenAI( base_url='https://external.api.recraft.ai/v1', api_key=, ) ``` # Pricing Source: https://www.recraft.ai/docs/api-reference/pricing ## API unit package pricing The following are the API Unit packages available from Recraft for the use of the Recraft API Service. API Unit packages must be purchased in advance and all API Unit packages are non-cancellable and non-refundable. Any number of unit packages can be bought, and purchased unit packages do not expire. | Price | API units | | ---------- | --------- | | USD \$1.00 | 1,000 | ## API unit charges for API services The following are the service charges (in API Units) for use of the API Services. API Units will be automatically deducted from Member’s pre-purchased API Unit package. | Service description | Cost (USD) | Cost (API units) | Billing Basis | | --------------------------------------------------------------------------------------------------------------------------- | ---------- | ---------------- | ------------- | | Raster image generation:
– Recraft V4.1 Pro
– Recraft V4.1 Utility Pro | \$0.21 | 210 | Per image | | Raster image generation:
– Recraft V4.1
– Recraft V4.1 Utility | \$0.035 | 35 | Per image | | Raster image generation – Recraft V4 Pro | \$0.25 | 250 | Per image | | Raster image generation – Recraft V4 | \$0.04 | 40 | Per image | | Raster image generation – Recraft V3 | \$0.04 | 40 | Per image | | Raster image generation – Recraft V2 | \$0.022 | 22 | Per image | | Vector image generation:
– Recraft V4.1 Pro Vector
– Recraft V4.1 Utility Pro Vector
– Recraft V4 Pro Vector | \$0.30 | 300 | Per image | | Vector image generation:
– Recraft V4.1 Vector
– Recraft V4.1 Utility Vector
– Recraft V4 Vector | \$0.08 | 80 | Per image | | Vector image generation – Recraft V3 Vector | \$0.08 | 80 | Per image | | Vector image generation – Recraft V2 Vector | \$0.044 | 44 | Per image | | Raster image to image – Recraft V3 | \$0.04 | 40 | Per image | | Vector image to image – Recraft V3 Vector | \$0.08 | 80 | Per image | | Raster image inpainting – Recraft V3 | \$0.04 | 40 | Per image | | Vector image inpainting – Recraft V3 Vector | \$0.08 | 80 | Per image | | Raster image outpainting – Recraft V3 | \$0.04 | 40 | Per image | | Vector image outpainting – Recraft V3 Vector | \$0.08 | 80 | Per image | | Replace raster background – Recraft V3 | \$0.04 | 40 | Per image | | Replace vector background – Recraft V3 Vector | \$0.08 | 80 | Per image | | Generate raster background – Recraft V3 | \$0.04 | 40 | Per image | | Generate vector background – Recraft V3 Vector | \$0.08 | 80 | Per image | | Image style creation | \$0.04 | 40 | Per request | | Image vectorization | \$0.01 | 10 | Per request | | Image background removal | \$0.01 | 10 | Per request | | Crisp upscale | \$0.004 | 4 | Per request | | Creative upscale | \$0.25 | 250 | Per request | | Erase region | \$0.002 | 2 | Per request | | Variate image | \$0.04 | 40 | Per request | | Prompt enhancement | \$0.01 | 10 | Per request | # Styles Source: https://www.recraft.ai/docs/api-reference/styles **Note:** Styles are **not yet supported for V4 models**. Styles define the visual appearance and aesthetic of generated images, including textures, visual effects, colors, composition, and overall artistic feel. Recraft offers both predefined curated styles and the ability to create custom styles. Recraft styles are organized into groups: realistic, digital illustrations, vector illustrations, icons, etc. Please refer to the examples below for visual references. To generate an image, you may provide a specific style name, such as `Recraft V3 Raw`, `Photorealism`, or `Vector art`. If a style name is supported by multiple models, the API defaults to Recraft V3 (if supported, to Recraft V2 otherwise). To use a specific version, explicitly pair the style with the model parameter (e.g., `Photorealism` with `recraftv3` or `Vector art` with `recraftv2_vector`). Refer to the [List of Styles](/docs/api-reference/styles#list-of-styles) for compatibility across V2 / V3 models. Examples: Images of style `Photorealism` are expected to look like just ordinary photographs made with a digital camera or a smartphone or a film camera. Photorealism Images of style `Illustration` are pictures drawn by hand or using computers - virtually everything except photos and vector illustrations. The most crucial difference from `Photorealism` is that illustrations possess simplified textures (like in 3D-rendered or manually drawn images) - or they are stylized in a certain creative way. The difference from `Vector art` is that `Illustration` allow for more complex color transitions, shades, fine textures. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b8318c6f6ecaf608531da_unicorn%20\(8\).png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82f54ef9617939df9d8e_motorcycle--going-fast.png) Images of style `Vector art` are expected to look like those drawn using vector graphics (see [Wikipedia](https://en.wikipedia.org/wiki/Vector_graphics)). Usually, they use only a few different colors at once, shapes are filled with flat colors or simple color gradients. Shapes of objects can be arbitrarily complex. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b82fb365f45ed1a743d43_a-man-working-in-his-workshop--standing-straight-a%20\(1\)photorealism-exp1-1.png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b830165533469004467d9_create-a-minimal-character-illustration-of-a-resea.png) Images of style `Icon` are small digital images or symbols used in the graphical user interface. They are designed to be simple and recognizable at small sizes, often visually summarizing the action or object they stand for, or they can act as the visual identity for an app or a website and are crucial in branding. ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b830d1720fd793595cf40_Group%204291.png) ![](https://cdn.prod.website-files.com/655727fe69827d9a402de12c/679b831e6ee14a2f933f112c_Group%204254.png) `Photorealism` and `Illustration` images are generated in raster formats such as PNG, WEBP, or JPG. These formats use pixels to represent details, making them ideal for photos and complex artwork. On the other hand, `Vector art` and `Icon` are generated in the SVG format. Unlike raster images, SVGs are made of mathematical paths and shapes, allowing them to scale to any size without losing quality. ### Using of styles The `style` parameter accepts the name of any curated style provided by Recraft. Examples include but not limited to: * Recraft V4.1, Recraft V4.1 Vector, Recraft V4.1 Pro, Recraft V4.1 Vector Pro: **styles are not supported** * Recraft V4.1 Utility, Recraft V4.1 Utility Vector, Recraft V4.1 Utility Pro, Recraft V4.1 Utility Vector Pro: **styles are not supported** * Recraft V4, Recraft V4 Vector, Recraft V4 Pro, Recraft V4 Vector Pro: **styles are not supported** * Recraft V3: [`Recraft V3 Raw`](https://www.recraft.ai/styles/b9af225b-6b98-43d0-a4b3-16343641e852), [`Photorealism`](https://www.recraft.ai/styles/6b98805f-9342-4d1e-9b53-bcb6b6d20ea1), [`Illustration`](https://www.recraft.ai/styles/3eb71874-5e13-4386-a74b-e42119186b0a), [`Enterprise`](https://www.recraft.ai/styles/e4eb5a3f-1b8b-4f5c-bada-bf6c2451cb17), [`Hand-drawn`](https://www.recraft.ai/styles/17027120-cf06-465f-a4d0-2e44ad8dee0d), [`Punk Graphic`](https://www.recraft.ai/styles/84789d26-1153-4faf-a1d1-c3e94d7a0035), etc * Recraft V3 Vector: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a), [`Line art`](https://www.recraft.ai/styles/19942671-ffa4-4867-9bbe-f5e4503dcc93), [`Engraving`](https://www.recraft.ai/styles/ce1f9161-a8dc-488f-9ff1-3f41082887b5), etc * Recraft V2: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1), [`Illustration`](https://www.recraft.ai/styles/b8772060-8802-4f3c-b051-73349b1017a5), [`3D render`](https://www.recraft.ai/styles/4659b2d6-a580-4287-849e-96190d7dbf30), [`Kawaii`](https://www.recraft.ai/styles/f8f1bc1c-1ae7-43ea-a36f-c81049feee01), etc * Recraft V2 Vector: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19), [`Icon`](https://www.recraft.ai/styles/645cb7ca-4be3-4d90-9f16-e9c26f6812f1), [`Doodle`](https://www.recraft.ai/styles/a2ba0c04-e40d-4210-8ed2-46d3537c546c), etc For the complete list of available styles organized by model and category, see the [List of Styles](/docs/api-reference/styles#list-of-styles). Some styles with identical names exist across multiple models (e.g., `Photorealism` is available in Recraft V2, and Recraft V3 models). Use the `model` parameter to explicitly specify which model version to use. The `style_id` parameter accepts: * style IDs created via the API; * style IDs from the Recraft web platform, accessible if : * you own the style; * the style is publicly available; * the style was explicitly shared to your account. To obtain a style ID from the web platform, open any style in the Styles panel and click the three-dot menu to copy the style ID. The `style` and `style_id` parameters cannot be used together. Specify one or the other, not both. If neither parameter is provided, the default style for the selected model will be used: * Recraft V3: [`Recraft V3 Raw`](https://www.recraft.ai/styles/b9af225b-6b98-43d0-a4b3-16343641e852); * Recraft V3 Vector: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a); * Recraft V2: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1); * Recraft V2 Vector: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19). Model compatibility: * custom styles created on the web platform are compatible with the model specified during creation; * custom styles created via the API are compatible with Recraft V3 and Recraft V3 Vector models only. ### List of styles To generate an image, you may provide a specific style name, such as `Recraft V3 Raw`, `Photorealism`, or `Vector art`. If a style name is supported by multiple models, the API defaults to Recraft V3 (if supported, to Recraft V2 otherwise). To use a specific version, explicitly pair the style with the model parameter (e.g., `Photorealism` with `recraftv3` or `Vector art` with `recraftv2_vector`). #### Recraft V3 * Photorealistic styles: [`Photorealism`](https://www.recraft.ai/styles/6b98805f-9342-4d1e-9b53-bcb6b6d20ea1), [`Enterprise`](https://www.recraft.ai/styles/e4eb5a3f-1b8b-4f5c-bada-bf6c2451cb17), [`Natural light`](https://www.recraft.ai/styles/90cf050a-5985-4ca8-b255-d370831701ca), [`Studio photo`](https://www.recraft.ai/styles/b8f16786-5794-47fd-b2b3-5f607d919d1c), [`HDR`](https://www.recraft.ai/styles/d0c61ed6-5f6c-4332-af27-e6a5fc2bde64), [`Hard flash`](https://www.recraft.ai/styles/41501ef8-26a6-4d1c-946c-d8b659ff11c3), [`Motion blur`](https://www.recraft.ai/styles/ca56e01b-a128-477f-b2f3-13a7191c8e91), [`Black & white`](https://www.recraft.ai/styles/7babd38f-dc87-4ccf-99de-d8834f8900f5), [`Evening light`](https://www.recraft.ai/styles/d92bef4c-605e-451d-a93f-a8d78f14effa), [`Faded Nostalgia`](https://www.recraft.ai/styles/889bee89-edce-4104-b572-df38c58e8659), [`Forest life`](https://www.recraft.ai/styles/51536681-7ee0-4604-bbb3-3a9e45e2b59f), [`Mystic Naturalism`](https://www.recraft.ai/styles/3ab91eeb-203f-4e00-865b-1e64c7854f08), [`Natural Tones`](https://www.recraft.ai/styles/45f34903-6da9-4468-966d-11314d845b41), [`Organic Calm`](https://www.recraft.ai/styles/06401905-a828-4652-a962-7026693c8cf7), [`Real-Life Glow`](https://www.recraft.ai/styles/336eabc6-da87-4a1e-a567-c3dd17bf5e79), [`Retro Realism`](https://www.recraft.ai/styles/2a4176f2-be17-49c5-b8b7-7a3bdd11a277), [`Retro Snapshot`](https://www.recraft.ai/styles/226190a0-5c43-4e0c-98a0-1a4ba63800cd), [`Urban Drama`](https://www.recraft.ai/styles/c61a727c-a9a8-40a5-a289-fdaf3a59eae1), [`Village Realism`](https://www.recraft.ai/styles/ba5e9c0c-4a20-492d-924d-a758b52b102d), [`Warm Folk`](https://www.recraft.ai/styles/db207d41-37bf-46ae-9e12-3201167c4cbe), [`Product photo`](https://www.recraft.ai/styles/57ed072c-3e0e-4df5-baba-83a6e87048b8); * Illustration styles: [`Illustration`](https://www.recraft.ai/styles/3eb71874-5e13-4386-a74b-e42119186b0a), [`Hand-drawn`](https://www.recraft.ai/styles/17027120-cf06-465f-a4d0-2e44ad8dee0d), [`Grain`](https://www.recraft.ai/styles/245b724f-8a24-4bde-aba4-21d41295074d), [`Bold Sketch`](https://www.recraft.ai/styles/6319466c-2bda-4b13-88ba-3e4e82aba55e), [`Pencil sketch`](https://www.recraft.ai/styles/8a5273c5-12cf-48de-8217-0f879b881398), [`Retro Pop`](https://www.recraft.ai/styles/a56d7265-ed27-4e0d-827b-2692d0b46fba), [`Clay`](https://www.recraft.ai/styles/f5145ada-d0d5-46d9-adda-d38fdf0b80d9), [`Risograph`](https://www.recraft.ai/styles/7be82d9c-ad40-483e-b235-389043675ad2), [`Color engraving`](https://www.recraft.ai/styles/75ee357c-ff80-4fbe-99ae-44d3310155fe), [`Pixel art`](https://www.recraft.ai/styles/aa32a571-eeeb-48bc-bc51-5792e9f49eb4), [`Antiquarian`](https://www.recraft.ai/styles/2b6d8781-7ab6-43dc-b870-891cc072246d), [`Bold fantasy`](https://www.recraft.ai/styles/1703ae77-b491-47e2-a25d-9dd6c28c2ffa), [`Child book`](https://www.recraft.ai/styles/ecd1de2b-02b7-43ce-9780-2ea91deb38f4), [`Cover`](https://www.recraft.ai/styles/6e403112-b52d-408b-8f78-539657229ade), [`Crosshatch`](https://www.recraft.ai/styles/7cccbbd2-c282-4ef6-a37f-889d1c503b74), [`Digital engraving`](https://www.recraft.ai/styles/f0999093-d0dc-4718-adba-a038a60ff6b2), [`Expressionism`](https://www.recraft.ai/styles/e2d8e964-ea62-43f4-a269-a47dd8728fe5), [`Freehand details`](https://www.recraft.ai/styles/eac001dc-5c63-4e6b-8d1f-390cbe5d2040), [`Grain 2.0`](https://www.recraft.ai/styles/824645a6-dc5b-4ad2-95c2-90bb6db5fbe1), [`Graphic intensity`](https://www.recraft.ai/styles/88b9e740-728c-4d00-9725-6d820859f6b7), [`Hard Comics`](https://www.recraft.ai/styles/98cef725-f770-498b-a90e-2eb4ab03cddb), [`Long shadow`](https://www.recraft.ai/styles/9a3edce0-424f-49c5-b372-61b0f589426c), [`Modern Folk`](https://www.recraft.ai/styles/6aa86245-e03f-432d-8b57-5fe5bfdb60cf), [`Multicolor`](https://www.recraft.ai/styles/da5cc27d-0389-4b5c-8a0d-fe5b76937214), [`Neon Calm`](https://www.recraft.ai/styles/1feb869d-edcc-4b6b-b6d0-413e9560f2dd), [`Noir`](https://www.recraft.ai/styles/956a5ac8-401f-4e5f-a034-ec458c371598), [`Nostalgic pastel`](https://www.recraft.ai/styles/4ce80d17-2e82-4f07-8ee7-c20931174da9), [`Outline details`](https://www.recraft.ai/styles/87cf0b6b-a495-4436-8598-897a71bf5db6), [`Pastel gradient`](https://www.recraft.ai/styles/9819278f-d0fe-4331-b7d4-37b0cf69150a), [`Pastel sketch`](https://www.recraft.ai/styles/149f13db-53ac-4c3e-8eb7-d5d44af04405), [`Pop art`](https://www.recraft.ai/styles/c5c3026e-66e5-412e-b87a-101512d745ac), [`Pop renaissance`](https://www.recraft.ai/styles/99c21c89-7de0-46dc-b447-e519be2c47f0), [`Street art`](https://www.recraft.ai/styles/c42616cd-f2c2-459c-a745-6f0ac86589d7), [`Tablet sketch`](https://www.recraft.ai/styles/c0c1b91e-8b46-4243-8f0b-fdea56c43701), [`Urban Glow`](https://www.recraft.ai/styles/acf7c4d0-ac61-4fb4-a3fa-d3a1fde6c014), [`Urban sketching`](https://www.recraft.ai/styles/ce069e6f-09bc-482b-9a05-fabf6658f27f), [`Young adult book`](https://www.recraft.ai/styles/e3838b40-dbac-450c-a06c-2c52b3b790c7), [`Young adult book 2`](https://www.recraft.ai/styles/b2449e5b-6954-453e-a168-c6e743dafd9a), [`Seamless Digital`](https://www.recraft.ai/styles/a054fca2-259e-449e-9b25-a7b2ec7a9aa3); * Emblem styles: [`Prestige Emblem`](https://www.recraft.ai/styles/1c2f96d5-c2cf-4311-909f-44190a8ea153), [`Pop Graphic`](https://www.recraft.ai/styles/a4681bb8-4e6b-47c6-ae61-d4dcdf9b9dbc), [`Stamp`](https://www.recraft.ai/styles/dac4f551-db80-4cdd-8778-8d6ad4fa992b), [`Punk Graphic`](https://www.recraft.ai/styles/84789d26-1153-4faf-a1d1-c3e94d7a0035), [`Vintage Emblem`](https://www.recraft.ai/styles/7faff2e0-bdce-4a2b-aea5-471f7fb88f5e); #### Recraft V3 Vector * Vector styles: [`Vector art`](https://www.recraft.ai/styles/773902ac-683a-4042-86a3-fe714eee345a), [`Line art`](https://www.recraft.ai/styles/19942671-ffa4-4867-9bbe-f5e4503dcc93), [`Linocut`](https://www.recraft.ai/styles/455a5cff-6d70-4d9a-945f-9256f279f50e), [`Color blobs`](https://www.recraft.ai/styles/7119b64b-b99d-4929-a418-3ead12ca1847), [`Engraving`](https://www.recraft.ai/styles/ce1f9161-a8dc-488f-9ff1-3f41082887b5), [`Bold stroke`](https://www.recraft.ai/styles/e33777c2-e1d3-4b16-bcdb-c7335e67a138), [`Chemistry`](https://www.recraft.ai/styles/e5aeb587-cac5-45ab-887f-bbc933a3d756), [`Colored stencil`](https://www.recraft.ai/styles/f5907d95-d032-4f70-a035-308bf8326ff3), [`Cosmics`](https://www.recraft.ai/styles/f14175ea-077f-4d8d-b1cb-e568ed906324), [`Cutout`](https://www.recraft.ai/styles/9ab9c49c-a934-4c83-af04-de6665a8d992), [`Depressive`](https://www.recraft.ai/styles/94b499fd-6408-48d1-8160-87661e93767c), [`Editorial`](https://www.recraft.ai/styles/59133ee8-2fc5-4225-bcd9-6934d21bcac8), [`Emotional flat`](https://www.recraft.ai/styles/782e2222-dbd8-46e0-bcec-1179d6c0e511), [`Marker outline`](https://www.recraft.ai/styles/1e018a0b-3299-445d-8812-06fd2f24eafa), [`Mosaic`](https://www.recraft.ai/styles/6c19cde9-52a6-40b3-9cc9-d81fe1fdd387), [`Naivector`](https://www.recraft.ai/styles/70192033-8db6-49af-b08b-f0e0043dddb8), [`Roundish flat`](https://www.recraft.ai/styles/857c6d5e-4039-45c1-8e31-12cfa25fa638), [`Segmented Colors`](https://www.recraft.ai/styles/8eafc64b-6483-46ca-858c-ef509ded10b5), [`Sharp contrast`](https://www.recraft.ai/styles/3ddad82f-d3c0-432e-ab7e-d9ed441a73c0), [`Thin`](https://www.recraft.ai/styles/f8300e4a-1b2f-4ad6-abbe-7e7180b87f30), [`Vector Photo`](https://www.recraft.ai/styles/b331f7bc-4ef6-4d17-a042-bae020cab91f), [`Vivid shapes`](https://www.recraft.ai/styles/85243d3e-dfd8-4f8b-83d7-456e72491814), [`Seamless Vector`](https://www.recraft.ai/styles/2e8f425e-b1a4-4c38-8f07-977bd8c2caf5); #### Recraft V2 * Photorealistic styles: [`Photorealism`](https://www.recraft.ai/styles/5a99ae89-3f14-433b-abb2-a1d8074b1ab1), [`Enterprise`](https://www.recraft.ai/styles/35ceb30e-50f5-46d4-9372-e95d8458a380), [`Natural light`](https://www.recraft.ai/styles/030886da-f6b9-4730-a1c1-4ce54a610a81), [`Studio photo`](https://www.recraft.ai/styles/0295270b-e050-45ad-8711-e29d1ac87d10), [`HDR`](https://www.recraft.ai/styles/07d689f1-ee16-49c4-a103-befb59bd7032), [`Hard flash`](https://www.recraft.ai/styles/4cb07726-d352-48cf-8f9b-b06eed4e0563), [`Motion blur`](https://www.recraft.ai/styles/cee8bee7-2085-4aca-91d5-2e3939f11dc8), [`Black & white`](https://www.recraft.ai/styles/6fefbfcf-9c93-4bce-8b31-bcbd86327d66), [`Product photo`](https://www.recraft.ai/styles/570c6fb3-1b01-4da0-afc9-0cd9997ffcc2); * Illustration styles: [`Illustration`](https://www.recraft.ai/styles/b8772060-8802-4f3c-b051-73349b1017a5), [`3D render`](https://www.recraft.ai/styles/4659b2d6-a580-4287-849e-96190d7dbf30), [`Glow`](https://www.recraft.ai/styles/077aeb2b-fca1-49e6-805b-6174bf55ee1a), [`Watercolor`](https://www.recraft.ai/styles/568daa6a-9631-44e7-a615-ce547146ae52), [`Hand-drawn`](https://www.recraft.ai/styles/a389a8b9-1209-463f-b816-0c08eb4f93c4), [`Kawaii`](https://www.recraft.ai/styles/f8f1bc1c-1ae7-43ea-a36f-c81049feee01), [`Grain`](https://www.recraft.ai/styles/9fe2962e-ac66-4d41-8ff4-38dbdfc79835), [`Bold Sketch`](https://www.recraft.ai/styles/41ef6ca6-8797-4cf1-87c6-522f425f2059), [`Pencil sketch`](https://www.recraft.ai/styles/f6e787d5-80cc-415e-b68d-860a0d25dafb), [`Retro Pop`](https://www.recraft.ai/styles/a28c4388-81fd-497c-a7ea-a44013636071), [`Clay`](https://www.recraft.ai/styles/c7aaa207-769f-4449-bdc4-c9d8cbcdbb09), [`Risograph`](https://www.recraft.ai/styles/902997d1-72a5-4951-81b8-042732f32d26), [`Psychedelic`](https://www.recraft.ai/styles/1f9a9d70-2d71-4b85-9106-646bbddf1fd0), [`Seamless Digital`](https://www.recraft.ai/styles/655b577f-a89f-4015-8a95-12b8daa2414b), [`Color engraving`](https://www.recraft.ai/styles/a36fc277-4f70-445b-bc1a-c50412c8c439), [`Pixel art`](https://www.recraft.ai/styles/7c8c1d97-17c8-44b4-a252-2808d44795a6), [`80's`](https://www.recraft.ai/styles/5e86f847-766c-4479-a6a0-645a54d77e06), [`Voxel art`](https://www.recraft.ai/styles/85770603-257a-4588-a9e7-73064991c1b7); #### Recraft V2 Vector * Vector styles: [`Vector art`](https://www.recraft.ai/styles/807ce9bc-7e36-4efa-a6cb-11cda1cecd19), [`Line art`](https://www.recraft.ai/styles/da8ce568-08d9-4bfb-9ef0-2984ec6391e3), [`Linocut`](https://www.recraft.ai/styles/3fc31d5e-a4ea-436c-be1a-787ec1cc76d8), [`Cartoon`](https://www.recraft.ai/styles/6cdae472-693d-478a-91fc-83eaf7c91a0c), [`Flat 2.0`](https://www.recraft.ai/styles/bcfabc0a-dec4-479b-962f-b306609e90ee), [`Color blobs`](https://www.recraft.ai/styles/2dc79001-283c-4c1e-ac01-49a95e21175c), [`Vector Kawaii`](https://www.recraft.ai/styles/6caa7c53-ad06-4d3b-827d-5bb81e782659), [`Doodle Line art`](https://www.recraft.ai/styles/1eb016d9-46d0-4b4e-bf1d-76296f43b471), [`Seamless Vector`](https://www.recraft.ai/styles/22e6d2f6-7951-49c4-aaaa-b7db63cb880a), [`Engraving`](https://www.recraft.ai/styles/100ae2cf-358a-46d2-ab13-dda7079c0460); * Icon styles: [`Icon`](https://www.recraft.ai/styles/645cb7ca-4be3-4d90-9f16-e9c26f6812f1), [`Outline`](https://www.recraft.ai/styles/5cf68b22-c903-48c5-882b-16872ef2d124), [`Pictogram`](https://www.recraft.ai/styles/be9ebb8f-312a-4d8e-b7ba-8276b05a2108), [`Colored outline`](https://www.recraft.ai/styles/7513cd33-5b71-4e04-8fed-becdd2672b4a), [`Doodle`](https://www.recraft.ai/styles/a2ba0c04-e40d-4210-8ed2-46d3537c546c), [`Colored shape`](https://www.recraft.ai/styles/1c5ca809-1bce-4b1c-8b14-79a5bc6df150), [`Gradient outline`](https://www.recraft.ai/styles/a7272add-907f-49f0-860d-d8b6d2dd0995), [`Offset doodle`](https://www.recraft.ai/styles/bdb513fc-cf7f-4e0f-be45-ef18bd693d03), [`Gradient shape`](https://www.recraft.ai/styles/65dcb522-d206-4561-bf93-9ae650f73136), [`Broken line`](https://www.recraft.ai/styles/d700cad3-94da-4204-b753-dbb75487edfc), [`Offset fill`](https://www.recraft.ai/styles/5a905231-9aee-4d4b-a0bb-eda7e79a0e95); # Swagger Source: https://www.recraft.ai/docs/api-reference/swagger OpenAPI is a standard format for describing RESTful APIs that allows those APIs to be integrated with tools for a wide variety of applications, including testing, client library generation, IDE integration, and more. To get started, check out the [Specification](https://external.api.recraft.ai/doc/#/). # Character consistency Source: https://www.recraft.ai/docs/best-practices/character-consistency Character consistency refers to maintaining a recognizable subject across multiple image generations. This can include not only people or mascots, but also products or objects—such as a specific chair, a branded package, or a hat. While Recraft doesn’t offer a dedicated character-tracking feature, you can achieve strong visual continuity using a combination of prompt engineering, style control, reference images, frames, and external model tools. ## What counts as a "character" * People (e.g., a red-haired woman in a space suit) * Mascots (e.g., a cartoon beaver used across marketing visuals) * Objects or products (e.g., the same pair of sunglasses shown in multiple lifestyle settings) ### Techniques for maintaining consistency * **Use a detailed prompt:** Start with a prompt that clearly defines your subject’s distinguishing traits. For example:\ *“Young woman with curly red hair, freckles, and an astronaut suit.”*\ This establishes a verbal template for regenerating or varying the character later.\ **Maintain a consistent style**: Apply the same style (or saved custom style) across all related images. Recraft interprets characters through the lens of visual style—keeping it consistent reduces unwanted shifts in appearance. * **Use frames to evolve or isolate specific traits**: Insert the character into a Frame to edit selected parts—such as changing clothes, expressions, or poses—without altering the full image. This is especially useful for storytelling or iterative design. * **Use reference images to preserve likeness**: Add a previously generated or uploaded image of your subject to the canvas and set it as a visual reference. Recraft will use this context to influence how new images are generated, helping preserve identity and structure. * **Use external models for remixing or variation**: You can attach one or more images (of a person, object, or mascot) to a prompt using the external model option. For example: * Combine two characters or items into a single scene. * Generate style or activity variations, such as:\ *“Make my selfie in the style of The Simpsons.”*\ *“Make me laughing.”*\ *“Make me riding a bike.”* These techniques can be layered together to iteratively build a series of images that feature a character or product with recognizable, coherent features, even as you explore different styles, scenes, or compositions. # Prompting and image generation Source: https://www.recraft.ai/docs/best-practices/prompting-and-image-generation * If the model isn’t interpreting your prompt as expected, try switching to a more general style, like **Recraft V3**, **Photorealism**, or **Illustration,** instead of a custom or niche style. * For logo designs with text that feel too rigid, enable **Negative prompt → Avoid text in prompt → Yes**. This can produce more expressive and creative text layouts, though it may introduce typos. * If prompt comprehension is poor, try lowering the **Artistic level**. This reduces creative variance and often improves adherence to prompt details. * If you're having trouble generating content that avoids the edges of the image, place the image in a **Frame** and use **Outpainting** to expand around it. * For small anatomical details (like faces or hands) that look distorted, run **Creative Upscale**. This often improves accuracy and visual quality. ### How to create mockups with designs If you’re working with a **complex surface** and need to experiment with how the design fits, such as adjusting size, placement, or angle, use **Convert to mockup**. If the mockup involves a **flat surface** where the design can be placed in a single, fixed way and doesn’t require experimenting with placement, use **prompt-based editing**.\ For example: a street banner or a phone screen. # Design Agent Source: https://www.recraft.ai/docs/design-agent ## From a Prompt to a Full Design Set Introducing the Recraft Design Agent A logo in one tool, an icon in another, a social post in a third. That's how most people building an app or a website end up with an identity: assembled, not designed. Fix one piece and the rest stops matching. The Design Agent skips that. Describe your product in a prompt, and it builds a full set of design assets: logo, palette, typography, app icon, icons, landing and hero visuals, social posts, and more, all pulling from the same creative direction. Save the result as a Design Kit, and every asset you pull from it afterward is built to match. Here's how it works. ### From prompt to Design Kit **Create a design set.** Give the agent a simple request, or a brief with references. It responds with four creative directions to choose from. Pick one, and the agent expands it into a full set, around 12 assets: logo, palette, typography, app icon, icons, a landing or hero visual, a social post, merch, and more. Key elements, like the logo mark or the palette, carry through across the set. **Save and reuse.** The finished set locks into a Design Kit. Come back next week for a new social post or a new icon, and the agent pulls from the same kit, so new assets are built to match what you already shipped. Edit or extend the kit as your product grows. No more re-prompting from scratch every time you need one more asset. ### **How to use it** There’s no new commands to learn. You talk to the Design Agent the way you'd brief a collaborator, right in the chat. It activates the moment your prompt calls for a coherent set: a brand, an identity, a kit of assets that need to hold together. For example: * "Help me create a brand kit for my app" * "Let's design a logo for my product idea" From there: 1. It reads your prompt and asks if you have references. 2. Four creative directions come back. Pick one. 3. Confirm which assets belong in the kit, and the Design Agent builds the full set, around 12 pieces: logo, palette, typography, app icon, icons, a landing or hero visual, a social post, merch, and more. The logo and palette carry through every asset. 4. It offers to save the set as a Design Kit. Add detail or ask for more directions at any point. Nothing here is final until you say so. ### **Saving and reusing a Design Kit** Once the design set ships, the agent saves it as a Design Kit. For now, a kit isn't shown as a separate item, it's stored as an agent “skill.” Reuse it anytime: open a new chat, ask for a new social post from your design kit, and the agent builds to match what's already out in the world. Extend the kit as the product grows. The logo you shipped in January still anchors what ships in June. ### Built for digital products The Design Agent works across styles and tasks, but this first release is tuned specifically for software: apps, websites, and similar digital products. That focus is deliberate. Narrowing the target lets us raise generation quality for the people who need it most right now: builders without a designer on the team, who need a coherent look fast and can't wait on five separate tools to agree with each other. ### Availability The Design Agent is available now in Recraft Studio, for subscribed users. It runs on credits: the more assets and iterations, the higher the cost, so a full project can add up. ### Start creating with the Design Agent Describe what you're building. See what one prompt can hold together. Start creating in Recraft Studio # Introduction Source: https://www.recraft.ai/docs/index Recraft has more than one model. Recraft's own models improve generation over generation. ## Recraft's model generations Recraft has released four generations of its own model: This is the latest generation. It delivers photorealism that actually looks real, sharper vector and illustration work, and strong results even from short prompts. It’s also the first generation to introduce Utility variants, built for flat, predictable outcomes when simplicity matters, making it perfect for product shots or mockups. This model is the previous generation and still a dependable default for everyday raster and vector generation. It's also the more opinionated of the two, leaning further into stylization than V4.1, with a knack for super stylish, unique photoshoots. The V3 model costs less per image than V4 and generates faster too. It's built for style consistency, precise text positioning, and artistic level control, and it's home to some of Recraft's most-loved styles, including Risograph and Hard flash. The vector variant is a strong pick for illustration work. This is Recraft's most affordable model. It's the one to reach for a specific look, like the long-loved Plastic 3D style, or when you need vector icons with consistent line width and shape across a whole set. ## Variants within a generation Most generations ship in more than one variant: * **Standard** is the default raster model. * **Pro** outputs at higher resolution, for print-ready assets and large-format work. * **Vector** generates editable SVG files directly, instead of raster images. * **Utility** (V4.1 only) uses flat lighting and front-facing composition, built for mockups and product shots rather than creative exploration. ## Not sure which to use? Start with [Choosing a model](https://www.recraft.ai/docs/recraft-models/choosing-a-model). It breaks down which model fits which job. # Getting started Source: https://www.recraft.ai/docs/mcp-reference/getting-started Recraft supports the [Model Context Protocol](https://modelcontextprotocol.io/) (MCP), enabling AI agents such as Claude, Cursor, and other MCP-compatible clients to interact with Recraft’s image generation and editing tools through a standardized interface. By connecting Recraft's MCP server to your client, you can generate and edit high-quality raster and vector images directly from your agent environment without switching tools. Recraft offers to use a remote MCP server at `https://mcp.recraft.ai/mcp` that requires no local installation. Authenticate via OAuth with your Recraft account and connect directly from any MCP-compatible client. Usage is billed against your Recraft **subscription credits** (the same balance used in the Recraft Studio app). See [MCP server](./remote-server) for setup instructions. ## Supported operations Recraft’s MCP server expose the following operations: * Raster and vector image generation * Raster and vector image transforms * Creating and using custom styles * Vectorization of raster images * Background removal and replacement * Upscaling of raster images See [MCP server](/docs/mcp-reference/remote-server) for setup details. ## Supported clients Any MCP-compatible client that supports the Streamable HTTP transport and OAuth can use the remote server. ## Troubleshooting * **OAuth window doesn't open or loops back.** Clear the cached OAuth state in your client and reconnect. For `mcp-remote`, delete `~/.mcp-auth/` (or the equivalent directory for your OS) and restart the client. * **Tools don't appear in the client.** Confirm the connector is enabled in your client's tools menu. In Claude Code, run `/mcp` to verify the server is connected and authenticated. * **"Insufficient credits" errors.** Top up your subscription credits on the [Recraft web platform](https://www.recraft.ai). MCP calls deduct from this balance, not from your API units. * **Generation succeeds but no image preview is shown.** Some clients impose a size limit on inline images. Retry with a smaller `size`, or open the returned URL directly. * **Tools call appears to hang.** Image operations can take up to a few minutes for large outputs or upscales. Wait for the call to time out before retrying. # MCP server Source: https://www.recraft.ai/docs/mcp-reference/remote-server Connect to Recraft's hosted MCP server without any local installation Recraft provides a hosted remote MCP server at `https://mcp.recraft.ai/mcp`. It supports the Streamable HTTP transport and uses OAuth 2.0 for authentication. The server exposes all functionality available through the [Recraft API](https://www.recraft.ai/docs/api-reference/endpoints), including image generation, editing, vectorization, upscaling, background removal and replacement, and custom style creation. ## Billing The MCP server consumes **subscription credits** from your Recraft plan — the same balance used by the Recraft Studio app. Each operation deducts credits according to the rates in [Plans and billing → Credits](/docs/plans-and-billing/credits). Free users receive a limited number of credits allowance; paid plans include a monthly allowance, with optional top-ups. ## Authentication The MCP server uses OAuth 2.0. When you connect for the first time, your MCP client will initiate an OAuth flow and redirect you to Recraft to authorize access. Once authorized, the client stores the token and handles authentication automatically on subsequent requests. ## Setup by client ### Claude App The Recraft MCP server is available as a custom connector in the [Claude App](https://claude.ai/) (web and desktop). The easiest way to install the connector is to follow this one-click link: [Add Recraft to Claude](https://claude.ai/customize/connectors?modal=add-custom-connector\&connectorName=Recraft\&connectorUrl=https%3A%2F%2Fmcp.recraft.ai%2Fmcp). The link opens the "Add custom connector" modal with the Recraft fields prefilled — review them and click **Add**. Alternatively, add the connector manually: 1. Open [Settings → Connectors](https://claude.ai/settings/connectors) in the Claude App. 2. Click **Add custom connector**. 3. Fill in the fields: * **Name:** `Recraft` * **MCP server URL:** `https://mcp.recraft.ai/mcp` 4. Click **Add** to save the connector. The first time you use the connector, Claude will redirect you to Recraft to authorize access via OAuth. Once authorized, the connector stays available across your Claude conversations — enable it from the **Search and tools** menu in the composer when you want Claude to use Recraft. ### Claude Code Run the following command in your terminal: ```bash theme={null} claude mcp add --transport http recraft https://mcp.recraft.ai/mcp ``` Then run `/mcp` in a session to authenticate via OAuth. **Allow image downloads on Claude Code on the web.** [Claude Code on the web](https://code.claude.com/docs/en/claude-code-on-the-web) runs in a managed container with restricted network access, so it may fail to download generated images from Recraft's image host. To fix this, allowlist the Recraft image host in the environment's **Network access** settings: 1. Open the environment for editing and find the **Network access** selector in the dialog. 2. Set the access level to **Custom**. 3. In the **Allowed domains** field, add the Recraft image host (one domain per line): ```text theme={null} img.recraft.ai ``` 4. Check **Also include default list of common package managers** to keep the default trusted domains alongside your custom entry. Once `img.recraft.ai` is allowed, Claude Code can download generated images without errors. ### Claude Desktop Open your `claude_desktop_config.json` file and add the following configuration: ```json theme={null} { "mcpServers": { "recraft": { "command": "npx", "args": [ "-y", "mcp-remote", "https://mcp.recraft.ai/mcp" ] } } } ``` On first use, `mcp-remote` will open a browser window to complete the OAuth flow. ### Cursor Open your `~/.cursor/mcp.json` file and add the following configuration: ```json theme={null} { "mcpServers": { "recraft": { "url": "https://mcp.recraft.ai/mcp" } } } ``` ### Visual Studio Code Open your `.vscode/mcp.json` file (or your user settings) and add: ```json theme={null} { "servers": { "recraft": { "type": "http", "url": "https://mcp.recraft.ai/mcp" } } } ``` ### Other MCP clients Any MCP-compatible client that supports Streamable HTTP transport and OAuth can connect using: * **URL:** `https://mcp.recraft.ai/mcp` * **Transport:** Streamable HTTP If your client does not support remote OAuth flows directly and requires a stdio-based command, use [`mcp-remote`](https://www.npmjs.com/package/mcp-remote): ```bash theme={null} npx -y mcp-remote https://mcp.recraft.ai/mcp ``` ## Supported operations The server supports the full set of image operations available through the Recraft API, including: * Generate raster and vector images from text prompts * Create image variations and image-to-image transformations * Edit selected parts of an image * Remove selected objects or regions * Remove, replace, or generate backgrounds * Convert raster images into vector graphics * Upscale images for either sharper or more detailed results * Create custom styles from reference images * View account details such as your remaining credits # Tools Source: https://www.recraft.ai/docs/mcp-reference/tools Recraft's MCP server allows use of the following tools. ## Billing The MCP server charges the same prices as the Recraft web platform and deducts from your **subscription credits** balance. See [Credits](/docs/plans-and-billing/credits) for credit costs per operation. ## Tools | Tool name | Description | Parameters | | -------------------- | ----------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------- | | `generate_image` | Generates raster/vector images from a prompt | - `prompt`
- `style`
- `size`
- `model`
- `number of images` | | `create_style` | Creates a style from a list of images | - `list of images`
- `basic style` | | `vectorize_image` | Vectorizes a raster image into a vector format | - `image` | | `image_to_image` | Generates raster/vector images from an image + prompt | - `image`
- `prompt`
- `similarity strength`
- `style`
- `size`
- `model`
- `number of images` | | `remove_background` | Removes the background of an image | - `image` | | `replace_background` | Generates a new background in an image from a prompt | - `image`
- `prompt for background`
- `style`
- `size`
- `model`
- `number of images` | | `crisp_upscale` | Upscales image resolution without altering content | - `image` | | `creative_upscale` | Upscales image resolution while regenerating details | - `image` | | `get_user` | Retrieves information about the user and balance | – | For a detailed explanation of each tool, its parameters, and pricing, visit the [Recraft API docs](https://www.recraft.ai/docs/api-reference/getting-started). ## Models and style compatibility Tools that accept a `model` parameter (`generate_image`, `image_to_image`, `replace_background`) work with any model from the Recraft V2, V3, V4, and V4.1 generations. A few compatibility rules apply: * **Styles are not supported on V4 and V4.1 models** (including their Pro, Utility, and Vector variants). The `style` and `style_id` parameters only apply to V2 and V3 models. See [Styles](/docs/api-reference/styles) for the full compatibility matrix. * **Vector models** (names ending in `_vector`) produce SVG output; raster models produce PNG/WEBP/JPG output. Pair your chosen model with a compatible style category — raster styles with raster models, vector styles with vector models. * **`custom_style_id`** created via `create_style` is bound to the model and base style supplied at creation time; reuse it with the same model. # Mobile Apps Source: https://www.recraft.ai/docs/mobile-apps Recraft is available on iOS and Android. Your account, projects, styles, and credits sync automatically across the mobile app and the web. Image 34 **Download the app** * [Download on the App Store](https://apps.apple.com/us/app/recraft/id6502416347) * [Get it on Google Play](https://play.google.com/store/apps/details?id=ai.recraft\&hl=en) **Requirements** iOS and iPadOS 17.6 or later. Android 8.0 or later. **Sign in** Open the app and sign in using the same method you use on the web — email, Google, or Apple. No password is required. If you sign in with email, you'll receive a one-time code to complete sign-in. Your credits and projects will be available immediately. **Recraft Studio on Mobile** Recraft Studio on mobile includes tools such as Edit Image, Upscale, Remove Background, and Vectorize. You can take a photo or upload an image directly from your phone to start working with these features. **Chat quick actions** The chat interface includes quick actions that allow you to retry the same generation or copy a message with a single tap. Available on both mobile and desktop versions. **Generate an image** Type a prompt in the prompt field and tap the generate button. Your image appears on screen in a few seconds. Tap any result to view, save, or continue editing. **Generation modes** Two modes are available: *Agentic mode* — An LLM-powered assistant understands context across multiple turns and optimizes settings automatically. Best for exploring ideas and iterating through conversation. *Manual mode* — Full control over every parameter, including model, style, aspect ratio, and artistic level. Best for precise, repeatable results. You can switch between modes at any time from the prompt panel. **Models and styles** The mobile app gives you access to all models available in Recraft Studio, including Recraft V4.1, Nano Banana, GPT Image, Seedream, and Flux. To select a model or style, tap the Style button above the prompt panel. **Reference images** You can attach images from your camera roll to guide generation. Tap the image icon in the prompt panel and select a photo. Use references to transform an existing image, build on a sketch, or match a visual style. **Video generation** Generate videos from a text prompt or from an image. Supported models include Kling, Veo, and Grok. Tap the video icon in the prompt panel to switch to video mode. You can control duration and keyframes for more precise results. Video generation is available to paid subscribers. **Continue on desktop** All images you generate on mobile are saved to your account and accessible from the History panel. Open any image on the desktop to continue working with the full toolkit, including the canvas editor, frames, mockups, and advanced editing tools. **Plans and credits** Your plan and credits are shared between the mobile app and the web version. Free users receive a limited number of credits. Paid plan credits reset monthly. You can purchase a paid plan or top up credits directly from the app. Top-up credits never expire. For more about plans, see [Paid plans](https://www.recraft.ai/docs/plans-and-billing/paid-plans). # Billing and invoices Source: https://www.recraft.ai/docs/plans-and-billing/billing Subscriptions can be billed on a **monthly** or **annual** basis. Annual plans include a discount compared to monthly billing. **Accepted payment methods** include credit cards, debit cards, and bank transfers. ## Canceling your subscription You can cancel your subscription at any time through the **Manage subscription** option in your profile. Your plan will remain active until the end of the current billing cycle. If you have used fewer than 30 credits over the lifetime of your account and your most recent payment was made within the past 30 days, you may request a refund. To do so, cancel your subscription and contact Recraft support at [help@recraft.ai](mailto:help@recraft.ai). Refunds are issued to the original payment method. ## Switching your billing cycle from yearly to monthly To change from yearly to monthly billing, you will need to first cancel your yearly subscription, then start a new subscription on a monthly billing cycle: 1. At the bottom left of the page, click on your name, then select **Manage subscription**. 2. Click **Cancel subscription** at the bottom, then complete the cancellation flow. 3. Go to the [Pricing page](https://www.recraft.ai/pricing) and click the toggle to select monthly billing. 4. Complete the purchase flow. Your yearly subscription will remain active until the end of your current billing period. After that, your plan will renew on a monthly basis. ## How to download an invoice or receipt 1. For personal accounts: Click on your profile avatar in the upper right, then select **Manage subscription**.
For teams accounts: Open the **Workspace** tab, then select **Manage subscription**. 2. Click **Manage billing details**. 3. In the Invoice history section at the bottom, click on your desired invoice. 4. Choose from **Download invoice** or **Download receipt**. # Commercial rights and ownership Source: https://www.recraft.ai/docs/plans-and-billing/commercial-rights-and-ownership Recraft’s usage rights depend on your subscription plan. Images generated under the **Free plan** are publicly visible in the community gallery and remain the property of Recraft. These images are not licensed for commercial use. For example, they typically cannot be listed on stock platforms, which require proof of ownership. Images generated under a **paid plan** grant you full ownership and commercial rights. You may use these assets for any purpose, including marketing, branding, product packaging, or other professional applications. Paid plan images remain private and under your ownership even after your subscription ends; they are not made public. Note: Regardless of plan, assets generated with Recraft may not be used to train any artificial intelligence models, systems, networks, or similar technologies. Learn more about image rights and ownership on [the Recraft blog](https://www.recraft.ai/blog/ownership-and-commercial-use-faq). # Credits Source: https://www.recraft.ai/docs/plans-and-billing/credits Recraft uses credits to power all image generation and editing operations. Paid plans include a monthly credit allowance, which resets at the start of each billing cycle. Credits are consumed based on the type of action performed. The table below summarizes the typical credit costs. The price depends on the selected model, and all prices are shown per image. Some operations have a base cost and are multiplied by the model price. \* **Multiplies by the selected model price.** For external models, please see [here](https://www.recraft.ai/docs/recraft-studio/image-generation/external-models#external-models). | Action | Credits | | | :---------------------------------------- | :------ | ------------------------------------------- | | Extract prompt | 1 | | | Creative upscale | 20 | | | Crisp upscale | 1 | | | Erase area | 1 | | | Generate a mockup | 2 | | | Modify area / Outpaint | 1\* | equals the model price (see Recraft Studio) | | Mockupize | 2 | | | Generate mockup | 2 | | | Generate image from prompt | 1\* | equals the model price (see Recraft Studio) | | Generate image set | 1\* | equals the model price (see Recraft Studio) | | Remove background | 1 | | | Vectorization | 1 | | | Remix | 1\* | equals the model price (see Recraft Studio) | | Replace background | 1\* | equals the model price (see Recraft Studio) | | Image Editing | 1\* | equals the model price (see Recraft Studio) | | Generate image variations | 1\* | equals the model price (see Recraft Studio) | | Exploration | 2 | | | Generate a video from a prompt | 1\* | equals the model price (see Recraft Studio) | | Generate a video from frames and a prompt | 1\* | equals the model price (see Recraft Studio) | | Upscale with external models | 1\* | equals the model price (see Recraft Studio) | | Prompt enhancer | 0 | | ## Credits in AI Chat mode In AI Chat mode, each chat message costs credits. The cost grows progressively as the conversation gets longer, because the model processes the full chat context with every new message. To reduce generation costs: * Start a new chat once the current conversation gets long. * Switch to Create mode, which doesn't use chat reasoning—you only pay for the generation itself. ## Credit types and expiration * Subscription credits reset monthly (or daily for Free users) and do not roll over. * Top-up credits never expire and remain available until used, even if the subscription ends. * When subscription credits are depleted, top-up credits are used automatically. * Credits can be earned through the referral program or purchased as needed (top-ups are available to paid users only). ## Credit top-ups Top-up credits can be purchased at any time, even before subscription credits are fully used. This ensures uninterrupted access to generation tools. It increments by 200, 400, 800, and 1600 credits. They never expire and are used after your monthly credits. and are only available to users on paid plans. **How to top up credits**: 1. Open Recraft and sign in. 2. Click on the credit counter icon in the upper right corner, then click **Buy more**. 3. A dialog box will appears to confirm your purchase, click **Add credits** to confirm. 4. Once payment is processed, the updated credit balance will appear (reload the page if necessary). Some models are provided by external providers and may have different credit costs. See the [External Models](https://www.recraft.ai/docs/recraft-studio/image-generation/external-models) page for more details. If assistance is needed, contact support at [help@recraft.ai](mailto:help@recraft.ai). Notes: * Top-up packs are available in fixed increments of 200, 400, 800, and 1600. * There is no limit to the number of top-ups that can be purchased. * Top-up credits are preserved across billing periods and subscriptions. # Deleting an account Source: https://www.recraft.ai/docs/plans-and-billing/deleting-an-account Deleting your account permanently removes your access to Recraft. According to our pricing policy, refunds are not available for deleted accounts. ### Desktop app 1. Log in to your account. 2. Click the **Profile** icon in the top-right corner to open the menu. 3. Select **Profile**, then choose **Delete account**. ### Mobile app 1. Tap your **Profile** icon in the top-right corner. 2. Select **Delete account** from the menu. # Free plan Source: https://www.recraft.ai/docs/plans-and-billing/free-plan The Free plan is ideal for trying out Recraft’s core functionality without a subscription. It offers limited credits, allowing you to explore Recraft’s image and vector generation tools along with basic editing features. You can upload up to three images per day, and each prompt produces up to two images at a time. Credit top-ups are only available on paid plans. Generated images in the Free plan are public and may appear in the community gallery. # Paid plans Source: https://www.recraft.ai/docs/plans-and-billing/paid-plans [***Refer to the Recraft pricing table for the latest credit allowances and subscription rates.***](https://www.recraft.ai/pricing) Recraft offers paid subscriptions for individual creators and teams with both monthly and annual billing options (annual billing provides up to 20% discount). For individual users, the **Basic plan** starts at: * **\$12 per month (billed monthly)** for 1,000 credits * **\$10 per month (billed annually, \$120 per year)** for 1,000 credits The **Pro plan** starts at: * **2000 credits** — \$20 per month (monthly) or \$16 per month (billed annually) * **4,000 credits** — \$40 per month (monthly) or \$32 per month (billed annually) * **8,000 credits** — \$80 per month (monthly) or \$64 per month (billed annually) * **16000 credits** — \$160 per month (monthly) or \$128 per month (billed annually) Credits renew monthly. Additionally, you can purchase top-up credits, which never expire and remain available until used, even if the subscription ends. For collaborative use, the **Teams plan** is available at: * **2000 credits** — \$22 per month (monthly) or \$18 per month (billed annually) * **4,000 credits** — \$44 per month (monthly) or \$35 per month (billed annually) * **8,000 credits** — \$70 per month (monthly) or \$88 per month (billed annually) * **16000 credits** — \$176 per month (monthly) or \$141 per month (billed annually) The Teams plan includes: * Shared custom styles * Centralized account management * Premium 24/7 support * Single Sign-On (SSO) All paid plans include private image generation and commercial usage rights while subscribed. ## Premium plans The Premium plans are intended for individuals who want to generate private images and retain [full ownership](/docs/plans-and-billing/commercial-rights-and-ownership). They include a monthly credit allowance and unlock access to all core editing and generation tools, as well as the ability to top up with additional credits as needed. Check out the full details on our [Pricing Page](https://recraft.ai/pricing) to see everything included! ## Credit top-ups You can also top up credit packs at any time, in increments of 200, 400, 800 and 1600. They never expire and are used after your monthly credits. Learn more about [credits](/docs/plans-and-billing/credits). ## Teams Recraft offers paid plans for teams and enterprise organizations. To learn more, visit the [Teams page](/docs/teams/teams-overview). # Upgrading and downgrading a plan Source: https://www.recraft.ai/docs/plans-and-billing/upgrading-and-downgrading-a-plan ## Upgrading Once confirmed, your plan will be upgraded immediately. Your billing will be prorated based on the remaining time on your current plan. **To upgrade to a paid Pro plan from a Free plan**: 1. Click **Upgrade** in the upper right screen. 2. Click **Get started** in the **Pro** plan. 3. Select the amount of credits you'd like to purchase, then click **Continue to payment**. 4. In the Stripe checkout window, confirm your new plan and complete the purchase. **To increase the credits on your current Pro plan**: 1. Open the **Profile menu** in the upper-right corner of the app. 2. Click **Pricing and plans**. 3. On the pricing page, locate the Pro plan in the center, then click **Current plan**. 4. Select the amount of credits you'd like to purchase then click **Continue to payment**. 5. In the Stripe checkout window, confirm your new plan and complete the purchase. ## Downgrading **To downgrade to the Free plan from a Pro plan**: 1. Open **Profile** in the upper-right corner of the app. 2. Click **Manage subscription**. In the Subscription box, click **Change plan**. 3. Click **Downgrade** under **Free**. 4. Click **Cancel subscription**. 5. Confirm changes. **To downgrade the credit limit from a Pro plan**: 1. Open **Profile** in the upper-right corner of the app. 2. Click **Manage subscription**. In the Subscription box, click **Change plan.** 3. Click **Current plan** under **Pro**. 4. Select the amount of credits you'd like to purchase, then click **Continue to payment**. 5. Confirm changes. After downgrading, you may receive a prorated amount of unused credits on your Stripe account that can be applied to future purchases. # Attributes Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/attributes Attributes shape the fine details of the image. They determine lighting behavior, color relationships, texture, materials, and typographic structure, adding polish without altering the core composition. Once the subject, composition, medium, and style are defined, attributes refine the surface qualities and atmosphere of the image. These cues help the model interpret nuance and realism while keeping the overall structure intact. **Lighting** Influences mood and focus, for example soft morning light, dramatic side lighting, diffuse overcast light, or harsh fluorescent light. **Color and palette** Shapes mood and visual identity, for example muted earth tones, high-contrast palette, pastel gradient, or bright complementary colors. **Texture and materials** Describe how surfaces interact with light such as matte metal, polished glass, rough linen, or grainy paper. **Typography and layout** (for graphic or vector work) Specify typographic hierarchy and structure such as bold sans-serif headline, subtle serif caption, minimal grid layout, or clean centered title. ## Examples
"*Portrait of a woman in a studio*" "*Portrait of a woman in studio with soft window light from left, gentle shadows, warm golden hour glow, natural diffused illumination*" "*Close-up portrait of woman in studio, face and shoulders filling frame, soft window light from left creating gentle shadows across features, muted earth tone palette, matte skin texture with natural detail, shallow depth of field f/2.8 with background softly blurred, slight film grain adding warmth, golden atmospheric glow, gentle rim lighting highlighting hair edges, natural elegant expression, contemporary editorial style, intimate framing, 85mm lens perspective*" # Medium Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/medium The medium establishes the visual language of the image. It defines the format or technique the model should follow and shapes how edges, textures, lighting, and materials are rendered. Specify the medium clearly, whether it’s a photograph, watercolor painting, 3D render, pencil sketch, or flat vector illustration. This helps the model understand the visual framework it should use when generating the image. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, natural window light, realistic details with water droplets, professional food photography, clean white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, watercolor painting, soft flowing edges, translucent washes, vibrant red and green palette, visible paper texture, white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, flat vector illustration, simple geometric shape, clean minimal design, solid red color with highlight, modern graphic style, white background*" # Style Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/style Style directs the model toward a specific artistic or visual tradition. It keeps the image coherent by aligning it with a clear aesthetic direction. Style can reference artistic movements such as Impressionism or Bauhaus, genres like editorial fashion photography or minimal poster design, or techniques like collage, ink sketch, or isometric render. Clear stylistic cues help the model stay focused on a consistent look rather than blending unrelated influences. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, authentic 1950s vintage advertisement style with aged paper texture, bold oversaturated red and green colors typical of mid-century print, prominent halftone dot screen pattern visible throughout, slight faded edges and subtle grain, retro commercial appeal with nostalgic quality, printed ephemera aesthetic, slight color separation misalignment for authentic vintage printing effect, classic Americana advertising look, warm sepia undertones, old magazine ad feeling*" "*Single whole red apple centered in frame, medium size filling one-third of image, Japanese minimalist aesthetic, zen simplicity, muted natural red tone, wabi-sabi philosophy, clean white empty space*" "*Single whole red apple centered in frame, medium size filling one-third of image, organic surrealist transformation with skin seamlessly morphing into liquid mercury on one side, roots and branches emerging from bottom creating impossible life cycle, butterflies made of apple flesh flying away leaving hollow spaces, soft pastel dreamlike background with melting clock elements à la Dalí, fruit existing in multiple states of matter simultaneously, poetic visual metaphor style, contemporary surrealist fine art photography, impossible beauty*" # Vibe Source: https://www.recraft.ai/docs/prompt-engineering-guide/artistic-principles/vibe Vibe sets the emotional quality of the image. It shapes mood and atmosphere so the output reflects your intended tone. Choose descriptors that clearly express the feeling you want to convey. Terms like serene, dramatic, playful, or tense help the model establish a consistent emotional direction. ## Examples
"*Single whole red apple centered in frame, medium size filling one-third of image, fresh morning harvest mood, crisp and inviting, bright natural light, healthy vibrant feeling, clean white background*" "*Single whole red apple centered in frame, medium size filling one-third of image, nostalgic autumn memory, warm golden tone, rustic countryside feeling, peaceful contemplative mood, soft neutral background*" "*Single whole red apple centered in frame, medium size filling one-third of image, dramatic moody atmosphere, dark black background with spotlight, intense shadows, theatrical chiaroscuro lighting, fine art still life*" # Closing notes Source: https://www.recraft.ai/docs/prompt-engineering-guide/closing-notes Prompting is a creative practice. The more you experiment, observe, and refine, the more natural it becomes to guide an image model with confidence. Each attempt teaches you something about how language shapes the result, revealing what matters most in your description and where a small adjustment can shift the entire mood or structure of an image. Start with what you see in your mind, describe it simply, and build from there. Some prompts will work immediately. Others will take a few iterations. Both are part of the process. Over time, you'll start to recognize patterns, develop preferences, and form a personal vocabulary that makes prompting feel intuitive rather than experimental. # Composition Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/composition Composition defines how the subject is positioned and viewed within the frame. Clear spatial instructions help the model understand perspective, balance, and overall framing. Specify how the subject should appear by describing camera angle, crop, pose, and perspective. A structured description helps the model interpret spatial relationships and produce a coherent image. ## Examples
"*Hands holding ceramic mug centered in frame, direct frontal view, cozy sweater sleeves visible, neutral background, soft even lighting, lifestyle close-up*" "*Hand resting beside coffee cup on left third of frame, full cup of coffee with steam on right side, wooden table surface, morning natural light, editorial beverage photography, minimalist composition*" "*Extreme close-up of hands kneading dough, tight crop on fingers and flour, texture details visible, warm directional light, macro detail shot*" # Context Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/context Context adds story, environment, and atmosphere. It helps the model understand the setting and the conditions that surround the subject. Provide supporting details such as environment, style, mood, or lighting to help anchor the scene. You do not need long descriptions, only enough to guide the model. Too little context leads to unpredictable results, while too much can restrict the outcome. Aim for a balanced level of detail that supports your intention without over-constraining the image. ## Examples
"*Full body shot of humanoid robot standing centered in frame, complete figure visible from head to feet, neutral white studio background, soft even lighting, clean product photography, frontal view, sleek modern design*" "*Full body shot of humanoid robot standing centered on foggy city street at night, complete figure visible, neon signs reflecting on metallic surface, rain-wet pavement around feet, cyberpunk atmosphere, cinematic photography, moody blue and pink tones, futuristic urban setting*" "*Full body shot of humanoid robot standing centered in overgrown abandoned greenhouse, complete figure visible from head to feet, dramatic contrast between advanced technology and wild nature reclaiming space, vines growing around base, soft natural light through broken glass roof, post-apocalyptic meets botanical, cinematic editorial photography*" # Subject Source: https://www.recraft.ai/docs/prompt-engineering-guide/core-principles/subject Start your prompt with the subject. This helps the model identify what the viewer should notice first and anchors the image around a clear focal point. Describe the primary character, object, or setting in simple, descriptive terms. Stating the subject clearly guides the model’s interpretation and keeps the output aligned with your intention. ## Examples Basic chair A wooden chair A detailed image of a wooden chair
"*A chair*" "*Wooden chair*" "*Mid-century modern walnut chair positioned beside tall window with afternoon sunlight casting diagonal shadows across seat, minimalist Scandinavian interior with white walls and light oak flooring, architectural photography style, clean composition with negative space*" # Level of detail Source: https://www.recraft.ai/docs/prompt-engineering-guide/depth-and-control/level-of-detail Detail level controls how much creative freedom the model has. By adjusting the amount of instruction you provide, you decide whether the model should explore broadly or render with precision. The level of detail in a prompt determines how much guidance the model receives and how closely it will follow your intention. Choosing the right amount of detail helps you balance predictability with creative exploration. Use level of detail intentionally. Expand it when you need specific results, and reduce it when you want variation and creative freedom. ## Examples
"*Avant-garde fashion portrait, sculptural headpiece*" "*Close-up avant-garde fashion portrait with dramatic white sculptural paper headpiece, model's face and shoulders visible, muted sage green background, soft studio lighting, editorial style*" "*Close-up editorial fashion portrait of model with natural beauty and no makeup wearing dramatic avant-garde white paper sculptural headpiece with geometric folded elements radiating outward, complete head and headpiece fully visible within frame from top to shoulders, face centered with breathing room around edges, natural skin with visible freckles and texture, serene calm expression with eyes looking directly at camera, wearing simple white high-neck garment, muted sage green studio background creating color harmony, soft diffused lighting from front with subtle shadows defining natural facial features, contemporary high-fashion meets authentic beauty aesthetic, architectural fashion photography, shot on 85mm lens at f/2.8 with gentle background blur, entire sculptural element captured without cropping, muted pastel palette emphasizing natural skin tones whites and soft greens, clean balanced composition, professional editorial photography celebrating raw natural beauty*" # Prompting for image generation Source: https://www.recraft.ai/docs/prompt-engineering-guide/introduction Cover image with several Recraft AI-generated images AI image generation allows you to create visuals using natural language instead of manual tools. With just a brief description, the model interprets your prompt and generates an image based on that input. For many users, the challenge is not producing an image, it is producing the right one. Even when your idea feels clear, the output can still look unexpected or misaligned with your intention. Effective prompting is not about memorizing magic keywords. It is about understanding how models interpret language and how simple adjustments to clarity and structure can guide the results. This guide introduces Recraft’s recommended prompting approach, built around a set of core components that help you communicate your creative idea more precisely. The **core principles** cover what every strong prompt needs at a minimum. The **artistic principles** are optional layers that enhance creativity, visual richness, and stylistic identity. Once these elements are clear, you can choose the image style you want to work in and decide whether your prompt should be brief or highly detailed. # Universal prompt template Source: https://www.recraft.ai/docs/prompt-engineering-guide/prompt-templates/universal This template brings together all prompt elements in a single structured format. It shows how subject, composition, context, medium, style, vibe, and attributes work together to guide the model toward a high-quality, coherent result. ## Template `[SUBJECT + ACTION], [COMPOSITION], [CONTEXT], [MEDIUM], [STYLE], [VIBE], [ATTRIBUTES: lighting, color, texture, technical details]` Prompt: A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward, captured from behind, dramatic mountain ranges stretching across the horizon in soft focus, endless field of blue and white flowers surrounding the robot, professional cinematic photograph, contemporary sci-fi style, contemplative and mystical mood, soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8 “A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward, captured from behind, dramatic mountain ranges stretching across the horizon in soft focus, endless field of blue and white flowers surrounding the robot, professional cinematic photograph, contemporary sci-fi style, contemplative and mystical mood, soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8” ## Breakdown 1. **Subject:** “A humanoid robot with black body standing in a vast wildflower meadow with arms raised upward” 2. **Composition:** “Captured from behind, dramatic mountain ranges stretching across the horizon in soft focus” 3. **Context:** “Endless field of blue and white flowers surrounding the robot” 4. **Medium:** “Professional cinematic photograph” 5. **Style:** “Contemporary sci-fi style” 6. **Vibe:** “Contemplative and mystical mood” 7. **Attributes:** “Soft diffused daylight with gentle atmospheric haze, cool color palette with pastel blues, whites, and muted greens, sharp focus on the robotic silhouette, slight depth of field blur on distant mountains, shot with 35mm lens at f/2.8” # Prompting with Recraft V4 Source: https://www.recraft.ai/docs/prompt-engineering-guide/prompting-with-recraft-v4 Image The model is designed to operate across different levels of control while consistently delivering outputs aligned with design-level quality and visual intent. You can describe intent briefly or define constraints in detail — both approaches are valid and produce stable results. * V4 accurately follows long, highly detailed prompts with multiple constraints. * V4 produces coherent, usable results from very short prompts. * V4 performs reliably in vector illustration and logo systems. * V4 generates stylized illustrations with clear silhouette, depth, and visual hierarchy. * V4 handles 3D visuals with consistent volume and material definition. * V4 works with text predictably, including multi-language text. * V4 delivers refined, designer-level photorealism when realism is required. In some cases, a short prompt is sufficient to explore form, mood, or composition. In others, longer prompts allow you to define structure, style systems, typography behavior, or production constraints.This guide focuses on practical usage: how prompt length, structure, and level of detail affect output — and how to choose the appropriate approach for different design tasks. ## Short Prompts → Interpretive Mode Recraft is capable of making informed aesthetic decisions when provided with minimal input. **Examples** Fashion couple portrait, close up. Fashion portrait, close up. Fashion man portrait, close up.
"*Fashion couple portrait, close up.*" "*Fashion portrait, close up.*" "Fashion man portrait, close up." ## Why this works: *Even minimal prompts produce:* * Balanced framing * Controlled lighting * Clean composition * Cohesive styling Recraft fills in missing structure using internal design logic. *When to use short prompts:* * Early-stage exploration * Mood discovery * Concept sketching * Allowing the model to introduce variation Short prompts activate interpretive behavior. ## Structured Prompts → Architectural Control If you want precision, define the visual system. **Examples** Centered close-up portrait of white woman. Artistic close-up portrait of a young Caucasian woman. Ultra-photorealistic high-angle bust portrait of an adult figure.
"*Centered close-up portrait of white woman, slightly closer than waist-up, natural three-quarter turn (semi-profile), wearing a structured tailored jacket with visible seam construction and precise cut lines. Matte fabric with realistic textile texture and subtle depth. Very natural-looking model with minimal to no makeup, authentic skin texture, soft imperfections visible.Clean seamless background without a horizon line. Natural daylight, soft and diffused, coming from the side (window light feel), gentle falloff across the face, delicate shadow under the chin and along the jawline. Calm, contemporary editorial mood. Balanced negative space, intimate framing, spatial clarity, quiet realism.*" "*Artistic close-up portrait of a young Caucasian woman and a pale horse against a dark desaturated blue-grey background with subtle vignetting. The woman stands on the right in sharp head-and-torso focus, slightly angled left, looking directly into the camera with a serious, neutral expression. Fair freckled skin, light reddish-brown hair pulled back, light eyes, no jewelry. She wears a high-neck metallic gold sequined garment over a dark brown layer, reflecting soft golden highlights.In the left foreground, slightly behind her, the pale horse’s head is softly out of focus but clear in form — white coat with faint speckles, light mane, visible eye, no tack. Strong natural light from the upper left creates sharp diagonal shadows and warm highlights against the cool background. Shallow depth of field, high contrast, eye-level composition, asymmetrical and intense mood.*" "Ultra-photorealistic high-angle bust portrait of an adult figure with dark skin visible at the neck and jawline, natural pores and realistic texture. Head slightly tilted downward, face fully concealed by a lightweight reddish-orange cotton head covering with visible weave, natural folds, and small light-blue and off-white floral print. A metallic silver crown of thorns wraps around the head, pressing into the fabric with realistic tension.Silver bead strands cross the forehead horizontally and hang vertically over the fabric and chest, each bead showing accurate reflections and subtle imperfections. The figure wears a deep navy corduroy kimono-style robe with ribbed texture and visible seams, with a cream underlayer at the neckline. Clear sky background fading from deep blue to pale haze over an arid landscape. Strong natural sunlight from the front-right, sharp shadows, realistic metal reflections, high dynamic range. High-end fashion editorial style, 8K clarity, solemn and contemplative mood." **Prompt structure (from global to local):** 1. Core concept — subject(s) and scene (who and what is in the image) 2. Background and environment (where the subjects exist) 3. Primary subject framing and pose (pose and expression) 4. Physical attributes and identity details (identity and appearance) 5. Secondary subjects and spatial relationships (if needed) 6. Lighting direction and behavior 7. Camera, depth, and contrast (how the scene is captured) 8. Mood and compositional resolution **Key takeaway:**\ Structured prompts don’t make results “better.”They make outcomes intentional, controllable, and repeatable. \ **Practical tip:**\ If the image must match a specific art direction → structure the prompt.If you’re exploring → keep it minimal. Image ## Vector & Logo Design Strength Recraft V4 is unusually strong at flat graphic logic. **Examples** Minimal playful logo. Minimal playful logo. Create a collection.
"*Minimal playful logo on a deep muted green background with warm off-white or cream elements. Flat colors only — no gradients, shadows, or texture. Designed for a slow, nature-focused creative space such as a studio, independent publishing project, or mindfulness platform. Calm, grounded, human, slightly whimsical — never commercial or trendy.Bold hand-drawn chunky wordmark with soft uneven strokes, rounded edges, imperfect spacing, and slightly irregular proportions. Letters feel brush-painted or cut from paper, earthy and intuitive rather than precise. Simple plant-inspired symbols (abstract leaves, berries, seeds) are integrated into or around the lettering — flat, bold, highly simplified shapes with no outlines or fine details.Centered, compact composition forming one unified silhouette with strong legibility. Warm, natural, quietly confident identity. Avoid corporate aesthetics, sharp geometry, thin lines, decorative fonts, or tech styling.*" "*Clean modern illustration icon set featuring exactly 4 ultra-bold surreal character pictograms in a playful contemporary mascot style. Exaggerated geometric bodies, elastic limbs, dynamic confident poses, slightly asymmetrical forms, smooth hand-drawn outlines with a subtle wobble. Minimal facial features — small oval eyes, calm neutral expressions. Personality expressed through posture. Strict two-tone palette only: deep saturated crimson for all characters and graphic elements on a soft warm cream background. Flat vector aesthetic. No gradients, no shading, no texture, no shadows. Include exactly four characters: – A tall melting heart with long flexible legs walking forward – A floating crescent-moon-headed figure hugging a large abstract flower – A stretched dachshund-like creature with an oversized bow around its body – A tall matchstick-headed character with a tiny flame-shaped head holding a large geometric love letter. Add minimal decorative elements in the same crimson line style: small starbursts, abstract leaves, curved motion lines, and floating petals. Clean, balanced, poster-style composition with strong negative space. No text. No clutter. Bold, graphic, collectible art-print aesthetic.*" "*Create a collection of 12 clean vector tourism character icons arranged in a grid with four per row (three rows total). Ultra-minimal bold silhouette style, pure black solid shapes on a white background. No gradients, no textures, no outlines — only simple geometric filled forms with strong silhouette clarity. All characters must have identical tiny dot eyes, same size and placement, with no additional facial details.The icons should feel playful, abstract, slightly absurd, and clearly tourism-inspired in a modern Scandinavian-brutalist pictogram style.Include exactly these 12 characters: a mountain with minimal sunrise rays, a rolling suitcase with small wheels, a tall lighthouse with beam lines, a rounded airplane, a palm tree with a face in the trunk, a camper van, a sun disk with short rays, a hot air balloon with a small basket, a backpack with visible straps, a triangular tent, a cruise ship with stacked decks, and a camera with a circular lens.Ensure even spacing, consistent visual weight, uniform eye placement, and a bold, clean, graphic poster-like composition.*" **Practical tip:**\ For logo and vector design, prompts should define: 1. Graphic type (logo, icon set, symbol system) 2. Shape logic (geometry, symmetry, silhouette clarity) 3. Color system (strict palette definition) 4. Line discipline (consistent stroke, no texture) 5. Layout structure (centered, grid-based, scalable) 6. Constraints (no gradients, no shadows, etc.) Avoid texture or material-focused language. Vector output responds to structural definition and geometric clarity. Image ## Graphic Design: Posters vol. 1 This is where Recraft V4 becomes especially powerful for designers.\ **It demonstrates a clear understanding of:** Typographic hierarchy\ Visual weight\ Poster-scale composition When layout logic is explicitly described, the model responds with accurate and predictable results. **Examples** Minimal Poster System Large-format poster. Large-format contemporary graphic design poster. Large-format contemporary magazine cover .
"*Large-format poster combining a cinematic desert landscape with bold graphic overlay. Background: open arid plain with warm sand tones, distant mountains, dramatic sky with soft evening light. Subtle print texture. Oversized electric-yellow abstract character drawn with thick expressive lines. A floating cloud-shaped figure with stretched limbs, one arm raised holding a symbolic lightning bolt. Simple facial features, closed eyes, calm expression. Flat solid fill, no gradients. Top left — minimal geometric logo mark + wordmark WILDFORM in clean modern sans-serif.Arched headline integrated with illustration: move slow move loud.Bottom line: Not perfect. Just alive. Strong contrast between realistic landscape and flat graphic layer. Contemporary, expressive, slightly rebellious poster aesthetic.*" "*Large-format contemporary graphic design poster combining raw editorial typography with expressive childlike crayon illustration. Warm off-white background with subtle paper grain and a strict grid layout defined by thin black frame lines and vertical divisions. At the top, an oversized tightly kerned bold black word in heavy grotesque sans-serif reads “BRAVE.” The center features a naive crayon drawing of a large sun and abstract landscape, made with thick wax strokes in saturated yellow, red, sky blue, and grass green, with uneven pressure and energetic scribbles. The drawing overlaps the grid and typography. Handwritten crayon text near it reads “dream big.”At the bottom, another oversized bold word reads “IMAGINE.” Between the illustration and lower headline, a narrow vertical column of small clean sans-serif text states: “Creative practice lives between structure and instinct. Design is not perfection — it is expression.” Footer in small uppercase sans-serif: “ARCHIVE / PRINT / POSTER / STUDIO / WORKS.”Brutalist black typography meets playful crayon art, flat colors, strong contrast, subtle paper texture — structured yet expressive and slightly rebellious.*" "*Large-format contemporary magazine cover combining bold naive painting with oversized modern typography. Background features a hand-painted still life of oversized seasonal vegetables — heirloom tomatoes, sliced purple cabbage, yellow squash, green zucchini, and red chili peppers — rendered in simplified flat shapes with thick matte acrylic texture and visible brushstrokes. The palette is saturated yet slightly muted, including dusty coral, mustard, sage green, deep plum, and soft cream tones. At the top, an oversized lowercase white wordmark in a heavy rounded sans-serif reads “harbor,” tightly spaced and spanning nearly the full width. Beneath it, a small clean sans-serif line reads: “design · culture · craft · studio · print · living.” A bold corner badge states “SPECIAL EDITION.” At the bottom, an organic cream-colored bubble contains the teaser text: “a handmade toolkit, slow projects, and a fold-out seasonal guide.”Clean editorial hierarchy, strong contrast between typography and painterly background, textured brush detail, modern independent print aesthetic — naive art combined with confident contemporary graphic design.*" **Prompt Structure:** 1. Format and scale (poster, cover, large-format) 2. Background / visual layer A 3. Graphic or image layer B 4. Typographic hierarchy (what is the largest?) 5. Text placement logic (top, center, edge, curved, cropped) 6. Contrast between layers 7. Overall compositional mechanics Recraft builds the layout instead of randomly placing elements. **Practical tip**: for hybrid posters (photo + graphic): Define each visual language separately, then describe how they interact. ## Graphic Design: Posters vol. 2 **Examples** Experimental Graphic Poster Playful glossy 3D poster. Promotional poster. Contemporary fashion campaign poster.
"*Playful glossy 3D poster featuring three iconic rubber ducks styled as high-fashion characters in a strong centered vertical stack. Each duck wears an exaggerated outfit: oversized futuristic sunglasses with a metallic puffer vest, a dramatic feathered collar with chunky jewelry, and a sporty streetwear hoodie with tiny sneakers. Smooth semi-gloss rubber texture, soft studio reflections, rounded toy-like proportions, subtle highlights, and confident minimal expressions. Background is flat saturated baby-blue. Layered 3D sticker elements float around the ducks — chunky stars, lightning bolts, smiley faces, safety pins, hearts, speech bubbles, and glossy tags in hot pink, yellow, cobalt, and white — creating diagonal movement and depth.Typography is bold and dynamic: inflated glossy hot-pink bubble letters reading “DUCK MODE” arch across the top, slightly overlapping the upper duck. A large diagonal shiny cobalt 3D headline “RECRAFT SPLASH” cuts across the lower half. A tilted oval sticker reads “LIMITED DROP.” Below the stack, small clean sans-serif text says “Designed for loud personalities.” In the corner, a glossy red capsule logo reads “QUACK CLUB — Tokyo / Milan.” High-fashion parody meets collectible toy culture, saturated, sculptural, poster-ready.*" "*Promotional poster for “System Lab” designer tableware collection from Formgrid Studio featuring sculptural ceramics and typography in a fresh grass-green tone — slightly acidic but not neon — set against a deep matte black background for sharp modern contrast. Centered symmetrical composition, photographed straight-on with a slightly high angle, soft diffused studio lighting from the top-front creating smooth highlights and gentle shadows.An extremely large bold condensed sans-serif headline “DESIGNER TABLEWARE” spans the full width at the top, edge-to-edge and slightly cropped, with smaller “VAJILLA DE DISEÑO” beneath. Oversized “SYSTEM LAB” overlays the center with smaller “LABORATORIO DE FORMAS” below. Vertical left text reads “EDITION 05 / EDICIÓN 05.” Bottom-right rounded rectangle contains “360° COMPOSITION / 360° COMPOSICIÓN,” with “FORMGRID STUDIO / ESTUDIO FORMGRID” nearby.The arrangement includes a sculptural low table with S-shaped legs and circular cutout shelf, an organic shallow platter with five petal-like lobes, a tall dual-handle glossy pitcher, and three matte bowls stacked in descending size. Minimalist industrial aesthetic, bold hierarchy, acid-leaning green on black, clean contemporary poster design.*" "*Contemporary fashion campaign poster blending realistic street photography with hyper-detailed 3D characters and bold experimental typography. A wide pedestrian crosswalk in a modern city under soft daylight — detailed asphalt, crisp zebra stripes, natural shadows, cinematic grading. A young avant-garde stylist walks confidently left to right, slightly off-center, captured in documentary realism with wind in the hair and a layered experimental outfit of saturated tones, technical fabrics, metallic accents, and structured tailoring.Plush-tech hybrid mascots move in the same direction along the stripes — one leading, one strong in the foreground, two behind at staggered depths — with ultra-real fur, glossy plastic and chrome details, and precise shadows aligned to the crosswalk. Seamless blend of real photography and CGI.Bold layered typography: “URBAN PARADE” oversized across the top, partially cropped; “NEW SEASON” and “COLLECTIVE DROP” as strong blocks on the right; date “04.12 — 05.08” in perspective along a stripe; curved tagline “Walk your story forward.” near the ground; compact “CITYFORM LAB” logo in the corner. Add large glossy sticker graphics — circular “EXCLUSIVE DROP,” rectangular “STREET EDITION,” diagonal “LIMITED RELEASE.” Strong asymmetry, unified motion, high-fashion meets collectible culture, dynamic editorial poster energy.*" **Prompt Structure:** 1. Core scene or subject 2. Additional 3D / graphic elements 3. Typographic dynamics (overlap, perspective, layering) 4. Color energy 5. Depth and spatial layering 6. Overall movement and tension ***Logic:*** Define layers first → then describe interaction. Image ## Illustration Strength Recraft V4 is highly capable in stylized and narrative illustration. **Examples** Digital anime-style. Сolorful illustration. Bold stylized illustration.
"*Digital anime-style chest-up portrait on a solid black background, centered. An androgynous young teen with cool pale skin and soft lavender blush, short choppy black hair with subtle teal highlights and small metallic clips. Large grey-blue eyes with bright highlights, thin straight brows, small muted rose lips, calm introspective expression.Wearing a modern sailor-inspired top in deep charcoal with a subtle geometric pattern, off-white collar trimmed in cobalt, and an oversized cobalt bow. Ten stylized koi fish in saturated reds, blues, gold, pink, white, and black swirl around the head and shoulders, overlapping the hair and partially crossing the face; one bold black-and-gold koi curves across the lower face. High contrast, flat lighting with soft shading, clean edges, dreamlike and contemporary mood.*" "*Сolorful illustration of a creative art studio filled with instruments and posters, but with a slightly more painterly, hand-drawn feel rather than clean vector graphics. A short-haired woman with visible arm tattoos and round glasses gently holds a large fluffy dog. She wears a loose white shirt and a dark hat. Around her are guitars, vinyl records, speakers, cables, sketchbooks, and framed portraits of abstract animal faces layered on the walls.Warm mustard, brown, terracotta, and burnt orange palette with soft tonal variation. Forms remain simplified and graphic, but edges are slightly irregular, as if drawn with ink and brush. Add a very subtle, delicate paper-like grain and gentle watercolor-style shading to avoid a flat vector look. Soft layered color transitions, light uneven brush marks, and faint depth in shadows. Cozy, artistic, intimate atmosphere — playful yet slightly textured, warm, and handcrafted rather than digitally perfect.*" "*Bold stylized illustration of an androgynous basketball player mid-dribble in extreme exaggerated perspective, with disproportionately enormous legs dominating the composition and oversized sneakers pushed dramatically toward the viewer. The entire figure is fully visible and fills the frame, but the legs occupy most of the visual space, making the torso appear smaller above them. Strong foreshortening and dynamic twisting posture.Color palette taken from the reference: saturated lemon-yellow ground, intense sky-blue background, vivid orange-red sun accent, bright acid neon green sneakers as the dominant highlight color, lavender-purple laces, soft beige midsole details, charcoal-grey outsole, crisp white socks, and bold black outlines. High-contrast flat graphic blocks.Add smooth gradients across the massive legs and shoes for volume, subtle airbrushed shading beneath the sneakers, and a delicate speckled grain overlay to avoid a clean vector look. Slightly irregular hand-drawn outlines remain visible. Energetic, surreal, contemporary poster aesthetic with bold acid green focus and exaggerated proportions.*" **Prompt Structure:** 1. Drawing style (anime, painterly, graphic, exaggerated) 2. Main character and pose 3. Line behavior (clean, irregular, bold) 4. Color logic 5. Surface treatment (flat, grain, watercolor shading) 6. Depth structure (gradients, airbrush, shadow softness) 7. Emotional tone \ ***Logic:*** You define drawing logic, not camera logic. Image ## 3D & Dimensional Visual Strength Recraft V4 handles dimensional form with strong material and lighting awareness. **Examples** Ultra-detailed stylized 3D render. Cinematic stylized 3D night. Digital 3D fashion lookbook render.
"*Ultra-detailed stylized 3D render in a high-end designer toy aesthetic from a top-down bird’s-eye view. A small sculptural character with rounded geometric anatomy and ultra-smooth semi-matte resin surfaces reclines diagonally on a large flowing chaise lounge with an integrated wicker detail, creating a strong graphic composition.The look is extravagant: lavender and powder-blue layered shoulders, dusty-coral torso, mint-green glossy knee-high boots, chunky pastel sneakers in mint, aqua and coral, plus oversized bangles, elongated earrings, a woven beige conical hat, tinted glasses, and a flowing ribbon accent. Around the chaise are a slate-blue side table, a peach ceramic vase with pastel branches, blush flowers, a folded parasol casting a radial shadow, and a woven rug.Soft peach, lavender, dusty coral, mint, wicker brown and blush tones dominate. Bright directional daylight creates long defined shadows from above, preserving a clean, sculptural, pastel-balanced editorial 3D still life.*" "*Cinematic stylized 3D night scene with strong depth and contrast, rendered in ultra-smooth plastic-like materials with refined matte surfaces and gentle subsurface scattering. A small stylized girl with an oversized head rides a skateboard through a dim urban alley. She has a playful, slightly mischievous expression, heavy-lidded minimal eyes, rounded proportions, two bouncing hair puffs, and a tiny snug beanie. She wears an oversized puffy jacket, loose wide-leg pants, and chunky sneakers in brighter electric blue and luminous cobalt tones, clearly separated from the darker surroundings.The background is deep and moody — navy-to-indigo gradient sky, shadowed brick walls, posters fading into darkness. The alley recedes with clear atmospheric perspective: sharper foreground, cooler midground, darker softly blurred background. Dry tiled pavement with subtle reflections, no rain or wet surfaces. Soft distant neon glow\.Lighting is cinematic: strong cool key light from above-right, soft side fill, and a clean rim light outlining her silhouette. Visible cast shadows from the skateboard and legs enhance depth. Low eye-level camera angle, shallow depth of field, modern animated short-film aesthetic with clear spatial layering and dimensional contrast.*" "Digital 3D fashion lookbook render featuring six anthropomorphic characters arranged in a clean 3 × 2 grid on a single solid light grey background and matching floor, figures slightly smaller in frame with generous negative space, soft frontal diffused lighting and gentle grounded shadows. All share identical minimalist faces with very small glossy bead-like black eyes, no pupils, no nose, and a thin minimal mouth line. Chibi proportions with oversized heads and compact bodies, matte plush materials with subtle metallic shine on layered chains. All styled as contemporary rappers with oversized silhouettes and bold streetwear attitude: a cream-white hare in an electric orange tee and sage varsity jacket with gold chains; a coral horned creature in a cropped white bomber over an apricot hoodie with beige cargos and a heavy chain; a charcoal raccoon in a moss oversized hoodie with layered chains and relaxed trousers; a lavender cat in an emerald oversized coat over a terracotta tee with wide charcoal pants; a black panther with subtle velvety fur wearing a cobalt hoodie and acid yellow crossbody bag with wide beige pants and platform sneakers; and a sand-colored fox in an ivory color-block jacket over a dusty-orange shirt with layered chains and white sneakers. Bright saturated palette, cohesive fashion-forward hip-hop collectible aesthetic." **Prompt Structure:** 1. Render type (designer toy, cinematic 3D, lookbook) 2. Form and proportion system 3. Material behavior (matte, gloss, plastic, fabric) 4. Spatial environment (floor, background, atmosphere) 5. Lighting direction and intensity 6. Camera angle and depth 7. Color system ***Logic:*** If you only describe the object, the result is interpretive.\ If you define the physical system, the result becomes controlled Image ## Working with Text  Recraft V4 handles text better than most generative systems, especially when you describe hierarchy. **Examples** Bold urban lifestyle. Bold avant-garde café menu. Contemporary premium food.
"*Bold urban lifestyle editorial featuring three avant-garde middle-aged men of distinctly different ethnic and visual backgrounds, with radically unique faces — varied bone structures, skin tones, wrinkles, scars, facial hair, asymmetry — real, unpolished, lived-in presence with no harmony or idealization. They sit close together at a small city street café table: one mid-bite holding food casually, one making a clear assertive hand gesture toward the camera, the third slightly back, calm and observant. All maintain direct or near-direct eye contact, confident, grounded, unapologetic.Shot at eye level in natural daylight with a documentary street-photography feel, tight framing on faces, hands, and table, honest skin texture and imperfections visible, softly blurred background. Controlled color palette with deep black base, clean white contrast, and one single bold accent color (cobalt blue, emerald green, or burnt orange), flat high-contrast tones only.Oversized experimental editorial typography overlaps the image without covering faces, using one or two short manifesto-style statements such as “NO RULES,” “CULTURE FIRST,” or “NOT FOR EVERYONE.” Multiple bold sticker-style graphic elements in the accent color visually connect the trio. Mood: raw contemporary street culture meets high-fashion editorial — confident, intelligent, manifesto-like, not made to please, not for everyone.*" "*Bold avant-garde café menu design in a luxury editorial style, photographed in an elevated, design-forward table setting. A large-format folded or partially unfolded menu card dominates the frame as the clear focal point, printed on thick premium matte paper with crisp edges and precise folds, placed on polished stone, lacquered wood, or tinted glass. The headline “VERA MENU” appears oversized and highly legible in an expressive high-fashion sans-serif with sculptural yet clean letterforms. The content is grammatically correct, clearly structured, and professionally written in proper English, with large evenly spaced typography and no filler text: sections such as ESPRESSO (Single-origin espresso, Double espresso), BREAKFAST (Sourdough toast with cultured butter, Soft scrambled eggs with herbs), and DESSERTS (Dark chocolate tart, Vanilla custard with seasonal fruit).Use a vivid high-impact palette — saturated coral, electric pink, deep cobalt blue, warm orange, and creamy off-white — with strong color blocking only, no gradients, and a disciplined typographic hierarchy. Surround the menu with minimal sculptural avant-garde cutlery featuring exaggerated forms and refined finishes. Soft directional editorial lighting creates subtle shadows while keeping the menu surface and typography crisp and fully readable. Overall impression: avant-garde luxury dining where graphic design meets contemporary art — bold, refined, collectible, no real brands, no clutter, focused on scale, clarity, and sculptural composition.*" "*Contemporary premium food packaging photographed in soft natural daylight, stacked rectangular matte cardboard boxes with sharp edges held slightly off-balance in human hands for a natural feel. Each box is a single refined color — deep forest green, warm cream, muted sky blue, soft clay beige — creating strong contrast. Bold oversized sans-serif brand name in dark ink is confidently cropped by the edges, with one short supporting line only, e.g., “VERA — Artisanal foods crafted slowly.” All typography is large, clear, and fully legible, no micro text or decorative clutter. A minimal abstract organic symbol (seed, leaf, or stone shape) appears flat and slightly irregular on the lid.One box is partially open, revealing two or three small clear glass jars with matte metal lids and simple bold labels containing gourmet spreads like pâté or eggplant caviar, rich natural textures visible. Matte paper texture with subtle grain, crisp printing, clean background, shallow depth of field. Modern, warm, premium editorial aesthetic — tactile, calm, design-forward, and quietly luxurious with intentional clarity in every element.*" **Prompt Structure:** 1. Format type (menu, editorial, packaging, poster) 2. Primary headline (size + placement) 3. Secondary hierarchy 4. Color blocking logic 5. Print surface and material 6. Spatial placement 7. Readability constraints 8. Place all required text in quotation marks ***Logic:*** Define the typographic system — not just the phrase. Quoted text ensures precise text rendering within the layout. Image ## Designer Photorealism Recraft’s realism feels curated rather than chaotic. **Examples** Ultra high-fashion. Cinematic fashion editorial. A cinematic editorial photograph.
"*Ultra high-fashion avant-garde close-up editorial portrait set in a cold cinematic studio. Deep monochrome cobalt-blue foggy background with a smooth gradient haze and soft atmospheric depth, minimal environment with no visible edges. The single character appears sculptural and iconic, with realistic natural skin texture — subtle pores, matte finish, sharp cheekbones, icy blue eyes, and an emotionless expression. Hair is platinum-blonde in a sharp geometric bob. They wear a dramatic couture piece featuring an oversized architectural ruffled collar made of translucent electric-blue tulle and layered organza, forming wave-like sculptural shapes around the neck. Makeup is minimal with muted burgundy lips and a soft powder finish.Strong long-exposure motion blur creates flowing trails and ghosted silhouettes behind the body and collar, while the face remains perfectly sharp. Clean, refined aesthetic with smooth gradients, no grain or painterly texture, high-end studio lighting. Cinematic, museum-like atmosphere with surreal quiet tension — modern couture captured mid-motion.*" "*Cinematic fashion editorial photograph captured outdoors during golden hour, focused on wind, motion, and atmosphere. The camera sits very low to the ground, almost inside tall wild grass, shooting upward to create an immersive, dynamic perspective. Softly blurred wheat-like plants dominate the foreground, forming organic diagonal lines that naturally frame the imageA young woman stands above the camera in a relaxed three-quarter pose, body slightly turned away while her head turns back toward the lens with direct eye contact, caught mid-movement rather than staged. She wears an oversized red hooded garment made of lightweight flowing nylon, voluminous and dramatic with visible drawstrings. Strong natural wind inflates and wraps the fabric around her body while her hair blows across her face, partially obscuring it.Warm golden-hour light casts a soft glow, with the sky fading from warm yellow near the horizon to gentle blue above. The setting is an open untamed field with no urban elements. Shallow depth of field keeps her face sharp while foreground grass and distant background remain blurred. Poetic, powerful, and alive — raw fashion captured in real wind, no studio lighting or artificial posing.*" "*A cinematic editorial photograph of a black man standing still on a busy crossing platform while crowds rush past her in heavy motion blur; he remains sharp and calm as people dissolve into streaks of movement; slow shutter effect, long exposure photography; muted urban color palette, soft industrial lighting, concrete textures, metro train in background; feeling of isolation, introspection, emotional contrast between chaos and stillness; contemporary fashion editorial mood, minimal styling, natural expression; shallow depth of field, cinematic framing, realism, high detail.*" **Practical tip:**\ \ Avoid stacking dramatic or evaluative adjectives.Precision and concrete description produce more reliable results than exaggeration.\ Prompt structure is flexible. The sequence of elements can be adjusted to change emphasis and influence model behavior. Elements placed earlier in the prompt receive higher priority and may shape the result more strongly.\ Common structural patterns include: * Scene / subject → style → details(prioritizes composition and content before aesthetic treatment) * Style → scene / subject → details(prioritizes visual language and overall aesthetic before composition) Both approaches are valid. Choose the structure based on which aspect should dominate the output. Image **How to Apply This in Real** **Branding:** Geometry → hierarchy → spacing → scalability → constraints **Fashion / Campaign:** Light → material → framing → simplified background **Vectors:** Silhouette → shape clarity → system consistency **Exploration:** 3–6 word prompts → generate variations **Posters:** Grid → margins → text size relationships **The Core Principle** **Recraft V4 adapts to your level of clarity** **|** Short prompts → model designs with you.\ Long prompts → model executes your architecture.\ \ **The model is strong in both modes |** Prompt engineering here is not about verbosity. \ It’s about defining structure, surface, space, hierarchy, and control.\ \ The more intentional your visual thinking, the more consistent and refined your results. # Graphic design Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/graphic-design Graphic design brings together images, text, and layout to communicate information clearly. Effective prompts emphasize structure, hierarchy, and consistent use of typography and color so the model can produce organized and readable compositions. ## Examples
"*Minimal poster design for jazz concert, bold yellow headline typography on black background, sans-serif lettering, high contrast*" "*Avant-garde surrealist magazine cover, model's portrait fragmented into floating mirror shards reflecting different dimensions, magazine title constructed from 3D letters dripping down image, impossible architecture emerging from model's hair, eyes replaced by cosmic nebulas, gravity-defying objects suspended in space, psychedelic color shifts, dreamlike Escher-inspired composition, cutting-edge experimental editorial design, museum-worthy surreal photography*" "*Bold 3D coffee advertisement, geometric low-poly coffee cup made of sharp triangular facets suspended in darkness, enormous brutalist sans-serif letters "DARK ROAST" dominating frame - typography so massive it bleeds off all edges, severe angular composition, monochrome black and espresso brown with electric orange geometric accent, Cinema 4D architectural rendering, minimalist maximalism, 65% oversized cropped typography 35% abstract 3D cup sculpture, contemporary luxury brand aesthetic, striking visual hierarchy*" # Illustration Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/illustration Illustration allows for expressive, stylized, or imaginative interpretation. The medium and emotional tone play a central role in shaping the final look. This style supports a wide range of techniques, from watercolor to ink to digital painting. Specify the medium and emotional tone to guide the model toward the character and atmosphere you want. ## Examples
"*Watercolor illustration of close up child and fox walking in the forest, soft diffused edges, flowing pigment washes, warm orange and green palette, visible paper texture, dreamy atmosphere, painterly style*" "*Black ink illustration of child and fox in forest, bold line work, crosshatching shadows, high contrast, pen and ink technique, detailed linework, white paper*" "*Digital painting of child and fox in forest, smooth brushstrokes, layered blending, soft lighting effects, contemporary illustration style, vibrant colors, polished finish*" # Logos and icons Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/logos-and-icons Logos and icons distill identity into simple, recognizable shapes that work at any scale. They depend on minimal detail, clear geometry, and a restrained palette. Direct prompts toward form and clarity to guide the model toward effective symbolic designs. ## Examples
"*Minimalist pasta brand logo centered in composition, circular icon with brand name integrated as negative space cutout within the circle, letters carved from pasta nest spiral, text visible through transparent areas of symbol, typography and icon fused as one shape, soft golden beige outline on white, modern serif letterforms becoming part of circular geometry, unified design where you cannot separate text from image, balanced central placement with white space buffer, zen minimalist aesthetic where form and meaning unite*" "*Vintage badge logo for artisan pasta brand, circular emblem with wheat sheaf and pasta shapes in center, ornate decorative border, ribbon banner with "Est. 1892", classic Italian typography, warm gold and cream palette, traditional family recipe seal design*" "*Line art icon logo for pasta brand, simple outline of spaghetti wrapped around fork, consistent stroke width, monoline style, clean and minimal, works at small sizes, modern culinary app aesthetic*" # Photorealism Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/photorealism Photorealism relies on accurate lighting, believable textures, and clear perspective. Emphasizing these elements helps the model produce images that resemble real photography. The goal is to make the image feel captured rather than created. ## Examples
"*Photorealistic medium portrait of woman in tall grass field at sunset, framing from shoulders up with face and flowing hair in focus, hands completely out of frame below, golden wheat and wildflowers surrounding her at shoulder level, warm golden hour backlighting creating rim light on hair, soft bokeh background with rolling fields, f/2.8 shallow depth, natural glowing skin, peaceful dreamy expression, simple natural style, warm amber and green tones, serene pastoral mood, lifestyle editorial photography, 50mm portrait lens, cinematic natural light*" "*High-end perfume product photography, elegant crystal bottle positioned on single geometric black obsidian block floating in void, dramatic spotlight from directly above creating sharp defined shadow beneath, rest in complete darkness, one precise beam of golden light hitting cap creating starburst reflection, ultra-minimalist composition with maximum negative space, luxury meets brutalist aesthetic, sharp focus throughout with f/11, museum-quality lighting setup, sophisticated contemporary commercial style, powerful visual restraint, Vogue-level editorial perfume shot*" "*Photorealistic wide shot of massive waterfall cascading into emerald pool deep in tropical rainforest, lush green vegetation covering every surface, morning mist rising from water creating ethereal atmosphere, sun rays penetrating through dense jungle canopy above, vibrant ferns and moss-covered rocks in foreground, crystal clear turquoise water, dramatic scale showing towering cliff walls, natural landscape photography, cinematic depth and layers, rich saturated greens and blues, shot on 16mm ultra*" # Vector art Source: https://www.recraft.ai/docs/prompt-engineering-guide/visual-formats/vector-art Vector art depends on precise shapes and clean geometry. Because it does not rely on texture or lighting for realism, prompts should highlight structure and simplified visual form. A focus on form, alignment, and clear edges guides the model toward the structured look typical of vector graphics. ## Examples
"*Flat vector illustration of mountain landscape, simple geometric shapes, two-color palette (blue, green), clean lines, minimal design*" "*Flat vector illustration of mountain landscape with layers, geometric peaks, stylized trees, gradient sky, bold outlines, modern graphic design style, balanced color palette*" "*Detailed flat vector illustration of mountain landscape, layered geometric mountains, stylized forest elements, gradient sunset sky, clean vector paths, contemporary poster design, rich color palette with complementary tones*" # Choosing a model Source: https://www.recraft.ai/docs/recraft-models/choosing-a-model **V4.1,** our latest model, delivers photorealism that actually looks real, sharper vector and illustration work, and stronger prompt understanding even from short prompts. Choose it when you want natural, lifelike photography, polished design and typography, or faster generation without sacrificing design taste. **V4** is the best default choice for most tasks. It delivers the highest design taste, prompt accuracy, and output quality across raster and vector. Use it for everyday generation and fast iteration. **V4 Pro** shares the same creative capabilities as V4 but outputs at higher resolution. Choose it when you need print-ready assets, large-format visuals, or pixel-dense detail. **V3** is a strong option when you need style consistency, precise text positioning, or artistic level control. **V2** remains valuable for specialized creative effects and unique style options that are exclusive to the model. # External Models in Recraft Source: https://www.recraft.ai/docs/recraft-models/external-models Recraft Studio gives you access to a curated selection of third-party image and video generation models alongside Recraft's own proprietary models. ## What are external models? External models are image and video generation models built by other providers that are available to use directly inside Recraft Studio. You can find them in the model selector alongside Recraft's native V2, V3, V4 and V4.1 models. ### Where can I use external models? External models are available in **Recraft Studio only**. Each generation consumes credits based on the model and quality setting. **External models are not available via the Recraft API.** The API supports only Recraft's proprietary model family (V2, V3, V4 and V4.1 generations). ### Supported external models for Image generation | Provider | Models | | :-------------------- | :------------------------------------------------------------------------------------------------------------------------------------ | | **OpenAI** | GPT Image 1.5 Low, Medium, High
GPT Image 2 Low, Medium, High | | **Google** | Nano Banana
Nano Banana 2
Nano Banana Pro
Nano Banana 2 4K
Nano Banana Pro 4K
Nano Banana 2 Lite | | **Black Forest Labs** | Flux 1 Schnell
Flux 1 Kontext Pro
Flux 1 Kontext Max
Flux 2 Dev
Flux 2 Pro
Flux 2 Max
Flux 2 Flex | | **Alibaba** | Qwen-Image 2
Qwen-Image 2 Pro
Z-Image Turbo | | **ByteDance** | Seedream V4
Seedream V4.5
Seedream 5.0 Lite,
Seedream 5.0 Pro | | **Ideogram** | Ideogram V4 Turbo
Ideogram V4 Balanced
Ideogram V4 Quality | | **ImagineArt** | ImagineArt 2.0 | | **Grok Imagine** | Grok Imagine Image | | **NVIDIA** | Cosmos 3 Super | | **Krea** | Krea 2 Large
Krea 2 Medium
Krea 2 Turbo | > *The list of available external models may change over time as new models are added or removed.* ## Credits and Reference Images Credit consumption varies depending on the selected model. The exact number of credits is displayed in the product when you choose a model. | Provider | Credits | Reference Images | | :-------------------- | :----------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **OpenAI** | 2–24 credits | Supports up to **10 reference images**. | | **Google** | 5–40 credits | **Nano Banana** models support up to **10 reference images**. | | **Black Forest Labs** | 1–18 credits | **FLUX.1 Kontext Max** and **FLUX.1 Kontext Pro** support **1 reference image**. **FLUX.2 Dev** supports **3 reference images**.
**FLUX.2 Pro**, **FLUX.2 Flex**, and **FLUX.2 Max** support up to **8 reference images**. | | **Alibaba** | 1–6 credits | **Qwen-Image 2**, and **Qwen-Image 2 Pro** support up to **10 reference images**. | | **ByteDance** | 7–22 credits | Supports up to **10 reference images**. | | **Ideogram** | 4–17 credits | Credit cost varies by model. | | **ImagineArt** | 8 credits | Credit cost is fixed at **8 credits**. | | **Grok Imagine** | 4 credits | Supports up to **3 reference images**. | | **NVIDIA** | 7 credits | Credit cost is fixed at **7 credits**. | | **Krea** | 2–8 credits | Credit cost varies by model. | > **Note:** The exact credit cost for each model is shown in the Recraft interface when you select it and may vary depending on the chosen model. ### Supported external models for Video generation | **Provider** | **Models** | | :--------------- | :--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **OpenAI** | **Sora 2**
**Sora 2 Pro** | | **Google** | **Gemini Omni Flash**
**Veo 3.1 Lite**
**Veo 3.1 Fast**
**Veo 3.1** | | **Kling** | **Kling 2.1 Master**
**Kling 2.5 Turbo**
**Kling 2.6 Pro**
**Kling 3 Turbo**
**Kling 3 Omni Standard**
**Kling 3 Omni Pro**
**Kling 3 Standard**
**Kling 3 Pro** | | **Pixverse** | **Pixverse V6**
**Pixverse V5.5**
**Pixverse V5.6** | | **Seedance** | **Seedance 1.5 Pro**
**Seedance 2.0 Fast**
**Seedance 2.0** | | **Alibaba** | **Wan 2.2 A14B**
**Wan 2.5 Preview**
**Wan 2.6**
**Wan 2.7**
**Happy Horse 1.0**
**Happy Horse 1.1** | | **Hailuo** | **Hailuo-02 Standard**
**Hailuo-02 Pro**
**Hailuo-2.3 Standard**
**Hailuo-2.3 Pro** | | **Ray** | **Ray 2 Flash**
**Ray 2** | | **Grok Imagine** | **Grok Imagine Video**
**Grok Imagine Video 1.5** | ## Video Generation Credits Video generation is priced **per second**. The total cost of a generation depends on both the **selected model** and additional video settings. In general, the longer the video, the higher the total credit cost. | Provider | Credits per Second | | :--------------- | :----------------------------------------------------------------------- | | **OpenAI** | 17–83 credits, depending on the selected model and additional settings | | **Google** | 13–142 credits, depending on the selected model and additional settings | | **Kling** | 12–56 credits, depending on the selected model and additional settings | | **PixVerse** | 7–80 credits, depending on the selected model and additional settings | | **Seedance** | 9–53 credits, depending on the selected model and additional settings | | **Alibaba** | 7–35 credits, depending on the selected model and additional settings | | **Hailuo** | 9–16 credits, depending on the selected model and additional settings | | **Ray** | 15–36 credits, depending on the selected model and additional settings | | **Grok Imagine** | 10-28 credits, depending on the selected model and additional settings | > **Note:** The exact credit cost is displayed in the Recraft interface when you select a model and configure your video generation settings. ### FAQ **Can I use GPT Image, Flux, or other external models through the Recraft API?** No. The Recraft API supports only Recraft's own models (V2, V3, V4 and V4.1 ). External models are available in Recraft Studio only. # Recraft V2 Source: https://www.recraft.ai/docs/recraft-models/recraft-V2 Recraft V2 is an image generation model released in March 2024 and the first model trained from scratch by Recraft. With 20 billion parameters, it was a breakthrough in human anatomical accuracy and the first to support brand consistency and brand color inputs. It also introduced vector image generation (SVG output), as well as minimalistic icon and illustration styles. V2 is available both in the web tool and via the API. In API use, it has a lower inference cost than V3. While V2 does not match V3 in prompt adherence or text accuracy, it remains preferred by many users for a set of specific styles and vector-based workflows. V2 is available in two models: | Service description | Cost (USD) | Cost (API units) | Billing Basis | Time | | :------------------ | :--------- | :--------------- | :------------ | :--- | | Recraft V2 | \$0.022 | 22 | Per image | 8s | | Recraft V2 Vector | \$0.044 | 44 | Per image | 14s | As of 2025, V2 is the only model that supports generating vector icons with brand consistency, allowing multiple icons to be produced with consistent line width, corner shapes, and other stylistic rules. ### **V2 curated styles** The curated library in V2 includes **Plastic 3D**, **Linocut**, **Grain**, and **Vector Art**, among others, and remain favorites of Recraft designers. * **Plastic 3D** produces consistently stylized 3D figures and icons in a chosen color palette, and is often used for app UI and presentation graphics. * **Linocut** is a vector style that can be monochrome or palette-based, and is frequently used for packaging. * **Grain** is valued for its textured look, common in blog and book illustrations. * **Vector Art** is popular for its distinctive outputs and is often used for logos, despite V2’s lack of accurate text generation. ### **V2 icon and pictogram styles** V2 includes several icon styles with different behavior. The **Recraft V2 Icon** style produces a different style for each generation when creating icons one-by-one. To achieve consistency, a new Vector Icon style must be created, or the Image Set tool can be used with **Recraft V2 Icon** selected. The **Recraft V2 Pictogram** style works similarly: each generation returns a new style unless the Image Set feature is used. This style also supports adjustable parameters such as cap type (see in-tool for available options). Other V2 icon styles produce consistent results by default. The most popular is **Vector Icon Outline**, which generates outline-style vector icons with a consistent look. ### **Limitations of V2** Some style names exist in both V2 and V3, but the visual results can differ significantly between them. V2 does not support accurate in-image text generation and may produce typographic errors. It uses the **Level of detail** parameter, which is not supported in V3, and does not support the **Artistic level** parameter available in V3. # Recraft V3 Source: https://www.recraft.ai/docs/recraft-models/recraft-V3 Recraft V3 is an image generation model released in **October 2024**. It ranked first for five consecutive months on the Artificial Analysis public benchmark on Hugging Face, maintaining a clear lead over models from Midjourney, OpenAI, and others. V3 is available in two models: | Service description | Cost (USD) | Cost (API units) | Billing Basis | Time | | :------------------ | :--------- | :--------------- | :------------ | ---- | | Recraft V3 | \$0.04 | 40 | Per image | 7s | | Recraft V3 Vector | \$0.08 | 80 | Per image | 16s | V3 introduced major advances in photorealism and text rendering. It was the first Recraft model to generate mid-size text accurately and, as of 2025, is the only model capable of placing text at specific positions in an image. It supports a rich library of curated styles, an infinite library of AI-generated styles, and like V2, can maintain style consistency from a small set of example images without retraining. V3 also includes the **Artistic level** parameter for adjusting the artistic fidelity of an image. Lower values produce more standard images, such as direct portraits, while higher values create less conventional compositions with more peculiar angles or perspectives. Because of its flexibility and quality, V3 is recommended for most use cases. # Recraft V4 Source: https://www.recraft.ai/docs/recraft-models/recraft-V4 Recraft V4 was released in **February 2026**. With a focus on visual and design taste, V4 produces images with balanced composition, cohesive color, and refined detail that feel intentional rather than stock-like, across both raster and vector outputs. Recraft V4 text-to-image models are available in four versions: | Service description | Cost (USD) | Cost (API units) | Billing Basis | Time | | :-------------------- | :--------- | :--------------- | :------------ | ---- | | Recraft V4 | \$0.04 | 40 | Per image | 10s | | Recraft V4 Pro | \$0.25 | 250 | Per image | 30s | | Recraft V4 Vector | \$0.08 | 80 | Per image | 15s | | Recraft V4 Pro Vector | \$0.30 | 300 | Per image | 45s | **Pro** delivers higher resolution and finer detail, perfecting anatomy issues and improving realism across complex scenes. It produces more reliable results for detailed compositions, print-ready visuals, and other high-fidelity applications. All versions of Recraft V4 are available on all plan types, including the Free plan. ## Exploration Mode **Exploration mode is temporarily unavailable.** We're re-imagining it to make it a better fit for your workflow — it'll be back soon. ## Key features **Design-forward visual judgment**: Produces images with balanced composition, cohesive color relationships, and controlled detail, resulting in a more stylish, modern look that can be applied directly in professional creative workflows. Outputs tend to feel deliberate rather than generic or stock-like, making V4 well suited for workflows where aesthetic quality matters alongside prompt accuracy. **Exceptional prompt understanding**: Interprets detailed creative direction, accurately following visual relationships such as reflections, materials, and background interactions. **Readable, structured text generation**: Produces clear, legible text for infographics, menus, signage, and packaging designs. Handles short and mid-length phrases with high fidelity to layout and typography. **Rich output variety**: Generate everything from stylized illustrations to photoreal images and production-ready vectors. V4 maintains visual coherence and high detail across diverse creative directions. **Sharper detail and higher realism**: Improved textures, color balance, and lighting deliver clean geometry and consistent visual depth, avoiding the flat or synthetic look common in stock-style imagery. **High-quality vector generation**: The only model capable of generating editable, production-quality **SVG** graphics. Vector outputs preserve scalable geometry and discrete color regions suitable for professional design workflows. **Higher native resolution (Pro version)**: Ideal for professional production workflows that require large formats or pixel-dense detail. The Pro model outputs at 2048x2048 px for design-grade clarity across print, product, and marketing use cases. **Supported formats** Export results in **SVG**, **PNG**, **JPG**, **PDF**, **TIFF**, and **Lottie**. **Default availability** All V4 models are available in the web app and via the API. ## Limitations The following features are not yet supported in Recraft V4: * Style creation * Prompt-based editing * Image sets * Artistic level control ## Availability Recraft V4 is accessible to all users via the **model selector** in the Recraft app and API, and replaces V3 for most generation tasks. ## Pricing Each image in exploration mode costs 2 credits. Since you always receive 8 images per generation, the total cost per generation is 16 credits. # Recraft V4.1 Source: https://www.recraft.ai/docs/recraft-models/recraft-v4-1 # More Beautiful by Nature Beautiful images don’t happen by accident. Neither does V4.1.

V4.1, released on May 14, 2026, makes images feel more alive, more purposeful, and more natural. The photorealism is quieter and more realistic. New illustration styles are now possible that simply weren’t before. And this model can read a short prompt and get the aesthetic right, all without the instructions of War and Peace to get there.

Recraft V4.1 text-to-image models are available in eight versions: | Service description | Cost (USD) | Cost (API units) | Billing Basis | Time | | :------------------------------ | :--------- | :--------------- | :------------ | ---- | | Recraft V4.1 | \$0.035 | 35 | Per image | 6.5s | | Recraft V4.1 Pro | \$0.21 | 210 | Per image | 12s | | Recraft V4.1 Utility | \$0.035 | 35 | Per image | 8.5s | | Recraft V4.1 Utility Pro | \$0.21 | 210 | Per image | 14s | | Recraft V4.1 Vector | \$0.08 | 80 | Per image | 12s | | Recraft V4.1 Vector Pro | \$0.30 | 300 | Per image | 15s | | Recraft V4.1 Utility Vector | \$0.08 | 80 | Per image | 14s | | Recraft V4.1 Utility Pro Vector | \$0.30 | 300 | Per image | 17s | The Pro models deliver higher resolution and finer detail, making them perfect model choices for large print-ready visuals and other projects. The Utility models are the ones to go for when simplicity matters more than creative surprise, making them perfect for mockups or product listing photos. The Vector models deliver fully exportable vectors that you can reshape, recolor, and tweak until they’re perfect. All versions of Recraft V4.1 are available to test on all plan types, including the Free plan. 6a05ad4d13f1ef510cf3e168 T Nv M Fw16 2 **Note:** Latencies vary between requests and depend on the user prompt and other parameters (e.g. aspect ratio). For some requests, latency can be lower than the values provided, and for others, it can be significantly higher. The values shown above represent median latencies. ## **Photorealism That Actually Looks Real** Screenshot 2026 05 14 At 22 18 34 > PROMPT
*Group selfie of friends standing close together, softly smiling with natural relaxed expressions, casual lifestyle photo, warm authentic atmosphere, modern youthful aesthetic, realistic lighting, candid social moment.* V4.1 understands people. In the group selfie above, it looks warm, diverse, and genuinely candid. There's less noise in the background. You can count on more natural photos, not just of the people themselves, but of the background as well. No more random urban backgrounds unless you specifically ask for it. Midjourney produces a beautiful photo here, but it delivers a lot more than "warm authentic atmosphere." The fuzzy, busy outdoor background has a life of its own. Screenshot 2026 05 14 At 22 19 42 > PROMPT
*Working on a laptop, desk setup, soft daylight* The difference holds up even in quieter settings. In the comparison above, the GPT Image 2 model generated a busier background that feels more like a stock photo than a real home. They might as well have put "Live, Laugh, Love" on one of those posters. The subject also has that slender, slightly Disney-coded face that the model defaults to. A nice picture? Sure. Realistic? Not quite. V4.1's image feels lived in. The desk is warm, a little messier, and real. V4.1 photorealism is built for images that need to feel human. It’s beautiful in the way that real life is beautiful. ## **Shorter Prompts, Better Results** Screenshot 2026 08 05 At 12 37 39 PM > PROMPT
*Couple tight hug* In this prompt, each model had only three words to get things right. V4.1 understood the assignment. The result is warm, intimate, and emotionally specific. It created a portrait photo of two people hugging with golden hour light that looks like a real couple photoshoot. Midjourney got the memo, but it went in a very different direction. It looks more like a first-person artistic view of a hug, not so much a photo of a couple in a tight hug. Screenshot 2026 05 14 At 22 22 12 > PROMPT
*Playing with a game controller, neon light*
V4.1 reads short prompts the way you'd imagine a creative director would. It fills the gap with taste instead of filling the frame with duplicates of things that no one asked for. With this prompt, Nano Banana 2 really goes all out with the "neon light" instructions, adding multiple neon signs to fill the background. Screenshot 2026 05 14 At 22 23 24 > PROMPT
*Wristwatch, metal and glass, close-up photo* In five words, Recraft read the room. The result is a sleek editorial shot with its dramatic angle and brushed metal catching light the way it does when a photographer with real taste is behind the camera. The competitor also produced a close-up photo of a wristwatch. On someone's wrist. On a park bench. Technically correct, aesthetically somewhere else entirely. Give V4.1 just a few words, and it'll give you something worth keeping. ## **Otherworldly 3D Design and Smoother Gradients** Screenshot 2026 05 14 At 22 24 17 V4.1 opens doors that weren’t there before. It renders 3D in a way that feels like it belongs in a gallery. Chrome catches light naturally. Compositions have real depth and weight. Screenshot 2026 05 14 At 22 24 43 The gradient work is in its own league too. Soft transitions, dreamy color blends, and surfaces that feel smooth rather than processed. It creates illustration styles that live in that space between digital art and something you could almost touch. Consider your creative studio officially expanded. ## **Vector and Illustration, Done Right** Screenshot 2026 05 14 At 22 25 31 > Prompt > > *Handwritten-script vector logo, single centered lockup, wordmark only. The word "Mirello" rendered in expressive brush-script lettering — fluid strokes, natural ink variation in stroke weight from thick downstrokes to hairline upstrokes, slight baseline bounce for a hand-lettered feel, generous swash on the opening "m" and a looping tail on the final "o" that underlines the full word. Closed letterforms with clean interior counters, no rough edges or texture. Strict two-color palette: charcoal black on green. Flat fills only — no gradients, no shadows, no drop effects, no outlines around the letterform. Compact horizontal lockup with even optical spacing. Scalable silhouette with strong legibility at small sizes. No decorative elements, no frames, no secondary text.* Text and illustration is where taste really shows. V4.1 has plenty of it. The V4.1 image is bolder, more expressive, and features a strong underline sweep that draws attention to the logo. The GPT Image 2 model is softer and more timid. The M loses its swagger and drama. GPT plays it safe. The letters are correct but lack conviction. It's the difference between a logo that commands attention and one that politely requests it. Screenshot 2026 05 14 At 22 26 47 > PROMPT
*Experimental typography spelling “kyro” made from fragmented sherpa fleece jacket panels, red, cream, green, and burgundy palette, sculptural textile letters with visible seams and piping, futuristic sportswear aesthetic, light-grey background, ultra detailed* Even when the prompt isn't as simple or straightforward, V4.1 makes sure that image is clean. With this kyro typography, the V4.1 model makes it sit flat on a light gray background, every letter structured and intentional. Nano Banana 2's version is textured and photographic, which looks impressive until you try to use it. It is plastered on a concrete wall at an angle and is difficult to understand. It doesn't look like something you'd bring to a exploration or brainstorming session. The V4.1 version, however, definitely belongs on a Pinterest board for someone's future fleece jacket business. Screenshot 2026 05 14 At 22 27 43 > Prompt
*A playful flat vector illustration of a skater captured mid-trick in a dynamic airborne pose against a solid saturated blue background. The character has exaggerated proportions and a bold stylized design, wearing a pink oversized T-shirt, loose brown pants, white-and-pink sneakers, and a small crossbody bag, while riding a bright yellow skateboard with simplified graphic details. One arm is raised holding a drink, the other balances the motion, and curved white swooshes wrap around the figure to emphasize movement, speed, and impact. The illustration uses clean shapes, minimal facial detail, smooth flat color fills, and a vibrant high-contrast palette that feels youthful, energetic, and contemporary. The overall image should feel fun, graphic, and full of personality, blending street culture, motion, and modern character illustration in a bold poster-like composition. Style notes: flat vector illustration, skateboarding character, exaggerated pose, bold color blocking, playful street aesthetic, clean shapes, dynamic motion graphics, contemporary poster feel.* The skater image is where V4.1’s illustration quality really shows its hand. V4.1 gives you bold color fills, precise shapes, and a composition that feels alive. Midjourney's version brings energy, but it loses the elegance with rougher lines and less consistent fills. One of these looks like a finished piece. The other looks like a first draft. Screenshot 2026 05 14 At 22 28 47 > PROMPT
*a hand-drawn character reference sheet with consistent identity* The character sheet is the most telling comparison of all. V4.1 creates a proper reference sheet with a front, profile, back, and 3/4 view, as well as adding character notes and color swatches. It looks like something an art student or anime fan would doodle. Midjourney draws a character. But Recraft brings one to life. ## **The Full V4.1 Family** ### **Recraft V4.1: Where It All Starts** V4.1 is the main model, and the most expressive one. It brings its own point of view to every prompt. It experiments with light, mood, and composition the way a good creative director would. It's built for exploration and concepting, for artistic self-expression, and for anyone who wants the model to bring something unexpected to the table. V4.1 is the model to choose when you need to create images that make people ask, "Who shot that?" Give it a direction and it'll take you somewhere beautiful. ### **V4.1 Vector: Sleek Illustrations and Text** V4.1 Vector is built for work where beauty lives in the line. Logos, typography, and illustrations are all crafted with the kind of precision and intention that makes the image feel created rather than generated. It understands lettering, it understands form, and it knows when something looks right. V4.1 Vector is the model to pick when you want full control over illustrations. ### **V4.1 Utility: Built for Simplicity** Screenshot 2026 05 14 At 22 29 45 Some projects don't need drama. They need clean, simple, and predictable. That's when you need Utility. Where the main V4.1 model experiments with angle, light, and composition, Utility focuses on flat lighting, front-facing composition, and simple scenes. It's the right tool when you're working on mockups, product shots, or anything where a calm, straightforward output serves the work better than an expressive one. ## **Start Creating with V4.1 in Recraft Studio** You've seen what it can do. Now it's your turn to see what one more pixel of progress looks like. Start creating via our [API](https://www.recraft.ai/docs/api-reference/getting-started) or in [Recraft Studio](https://www.recraft.ai/auth/login). # Recraft Plugins Source: https://www.recraft.ai/docs/recraft-plugins Download Recraft integrates directly into the tools you already use. Through native plugins and browser extensions, you can generate images, vectors, icons, and more without leaving your design or production environment. All plugins connect to your Recraft account and draw from your existing credits and plan. ## **Recraft for Figma**
The [Recraft Figma](https://www.figma.com/community/plugin/1613917155067897575/ai-image-generator-recraft) plugin brings AI image and vector generation directly into your Figma workspace. Generate raster images or fully editable SVG vectors without switching tools, and apply editing operations on any image inside your file. ***Features*** • **AI image generation** - Generate raster images from a text prompt using Recraft V3, V4, or V4 Vector Pro models. • **Vector generation** - Produce editable SVG vectors, including icons and logos, ready to use as Figma layers. • **Vectorization** - Convert any raster image to a clean, scalable vector. • **Background removal** - Remove image backgrounds in one click. • **Upscaling** -  Increase resolution for print or high-density screens. • **Templates** — Browse Recraft's template library to guide style and composition. ***How to install*** • Go to the Recraft plugin page on [Figma Community](https://www.figma.com/community/plugin/1613917155067897575/ai-image-generator-recraft). • Click **Open in Figma** — the plugin is added to your account automatically. • In any Figma file, open **Plugins** → **Recraft**. • Sign in with your Recraft account. ***How to use*** • Open the plugin from the Plugins menu in any Figma file. • Enter a prompt describing the image or vector you need. • Select a model (V3, V4, or V4 Vector Pro) and any style preferences. • Click **Generate**. • Drag the result onto your canvas, or use the editing tools to refine it. ***Typical workflow*** A product designer generating UI illustration assets can stay in Figma throughout: write a prompt, generate an SVG icon, drag it to the canvas, and adjust it like any other vector layer. Background removal and upscaling are available on selected images without leaving the file. ## **Recraft for Framer**
The Recraft plugin for Framer lets you generate images and vectors for web projects directly from the Framer canvas. It is designed for website and landing-page workflows where visual assets need to be web-ready and optimised for performance. ***Features*** • **AI image generation** — Generate raster images using Recraft V3, V4, or V4 Vector Pro. • **Vector generation** — Create editable SVG illustrations, icons, and logos. • **Vectorization** — Convert uploaded raster images to vector format. • **Background removal** — Clean up images for transparent-background use in web layouts. • **Upscaling** — Increase resolution for hero images and full-width sections. • **Templates** — Browse templates to keep style consistent across generated assets.
***How to install*** No installation required. Access the plugin directly from the Framer canvas: • Open a Framer project and click browse all **Plugins** icon in the toolbar. • Search for **Recraft** and click **Open Plugin**. • Alternatively, visit the Recraft page on the [Framer Marketplace](https://www.framer.com/marketplace/plugins/recraft/). • Sign in with your Recraft account. ***Practical examples*** • **Hero section illustration** - Generate a custom illustration from a prompt and drop it into the hero section of a landing page. • **Icon set** - Use V4 Vector Pro to generate a consistent set of SVG icons for a navigation or feature grid. • **Background image** - Generate a full-width raster image, upscale it for high-resolution displays, and set it as a section background. ## **Recraft Chrome Extension** The [Recraft Chrome Extension](https://www.recraft.ai/recraft-integrations/chrome-extension) AI image generation and editing to any webpage. It opens as a side panel and works alongside Google Docs, Google Slides, Gmail, and other web-based tools, letting you generate and insert assets without leaving the browser tab. ***Features*** • **Image generation on any webpage** - Generate raster images and vectors from a text prompt in the side panel. • **Image editing** - Pick any image on the current page to edit, or upload your own. Apply background removal, vectorization, and upscaling. • **Direct insertion** - Insert generated images directly into Google Docs, Google Slides, and Gmail. • **Templates** - Browse and apply Recraft templates from within the extension. ***How to install*** • Visit the Recraft extension page on the Chrome Web Store. • Click **Add to Chrome**. • Pin the Recraft icon to your browser toolbar for quick access. • Click the icon on any webpage and sign in with your Recraft account. ***Typical use cases*** • **Google Docs** - Generate an illustration mid-document and insert it at the cursor position. • **Google Slides** - Create a custom image for a slide without opening a separate design tool. • **Gmail** - Add a generated image to an email directly from the side panel. • **General browsing** - Click any image on a webpage to remove its background or upscale it. ## **Recraft for Google Docs** The [Recraft Google Docs](https://www.recraft.ai/recraft-integrations/google-docs) lets you generate raster images and insert them directly into a document. You can also edit images already in the document using background removal and upscaling. *Vector image generation requires opening Recraft directly at [recraft.ai](http://recraft.ai).* ***How to install*** • In Google Docs, go to **Extensions** → **Add-ons** → **Get add-ons**. • Search for **Recraft** and click **Install**. • Grant the required permissions and sign in with your Recraft account. • Access the add-on via **Extensions** → **Recraft**. ## **Recraft for Google Slides** The [Recraft Google Slides](https://www.recraft.ai/recraft-integrations/google-slides) add-on works the same way as the Docs add-on, but outputs images directly onto your slide. Background removal and upscaling are supported for images already on the slide. *Vector image generation requires opening Recraft directly at [recraft.ai](http://recraft.ai).* ***How to install*** • In Google Slides, go to **Extensions** → **Add-ons** → **Get add-ons**. • Search for **Recraft** and click **Install**. • Grant the required permissions and sign in with your Recraft account. • Access the add-on via **Extensions** → **Recraft**. ***Common Features Across Plugins*** All Recraft integrations share the same underlying models, editing tools, and account system. The table below shows feature availability per platform. Please note that for vectorization and vector images, you need to go to Recraft if you are using Google add-ons. Image **Models** • **Recraft V3** — Fast, reliable generation for standard workflows. • **Recraft V4** — Latest raster model with improved prompt understanding and built-in design taste. • **Recraft V4 Vector Pro** — Produces editable SVG vectors, icons, and logos. ***Credits and plans*** All plugins share the same credit pool as your Recraft account. Free users receive a limited number of credits. Additional credits and advanced features are available on paid plans. ***Best Practices*** **Use templates as style references** Templates help maintain visual consistency across a project. Apply a template before generating to anchor the style, then adjust with your own prompt. **Choose the right output format** Use vector output (V4 Vector Pro) for icons, logos, and UI elements that need to scale or be edited at the layer level. Use raster output (V4) for photographic-style images, textures, and backgrounds. **Upscale as a final step** The upscaling tool is most effective after composition is finalised, not mid-workflow. Run it once on the completed image before exporting. **Write specific, concrete prompts** Prompts that describe subject, style, and context together produce more consistent results than short or abstract prompts. Example: 'flat vector icon of a leaf, minimal, two colours, dark green on white'. **Keep styles consistent across a project** Use the same model, template, and style settings across all generations in a project. Switching models mid-project typically produces visible inconsistencies. **Limitations and Notes** • **Vector generation in Google tools** - Not available in Google Docs or Slides add-ons. Use [recraft.ai](http://recraft.ai), the Figma plugin, or the Framer plugin for vector output. • **Credit usage** - Each generation, upscale, or background removal consumes credits. Credit cost depends on the operation and model used. • **Browser support** - The Chrome Extension requires Google Chrome or a Chromium-based browser. Not available for Firefox or Safari. • **Export format** - Figma and Framer plugins output within the platform's native layer system. Export formats depend on each platform's own export settings. • **Feature parity** - New capabilities are released to [recraft.ai](http://recraft.ai) first and rolled out to plugins over time. • **Authentication** - Each plugin requires a sign-in with your Recraft account. All plugins share the same account credits and settings once connected. # How to create a color palette Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/how-to-create-a-color-palette You can create a custom palette in one of three ways: 1. **Add hex codes manually**\ Enter specific color codes from your brand guide or visual spec to control the palette with precision. 2. **Use the eyedropper tool**\ Sample colors from any image on the canvas using the color picker. This is helpful when referencing past generations or imported brand materials. 3. **Extract colors from an image**\ Upload an image from your computer and Recraft will automatically extract the dominant colors. This is ideal for building palettes based on moodboards, campaign assets, or photography. Palettes are saved to your library and can be edited, renamed, or deleted at any time. A small color indicator on the palette icon shows whether a palette is currently active. # Recoloring existing images Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/recoloring-existing-images ## Raster images For raster images, you can’t change specific colors directly. However, you can adjust overall tone and feel by selecting **Adjust colors** in the context panel located on the right of the canvas. This gives you access to sliders for: * Color spectrum * Saturation * Brightness * Contrast * Opacity Recoloring is especially helpful for light retouching or shifting the mood of a generated photo or painting. ## Vector images Vector images offer more detailed color control. After generation, an **Adjust colors** options appears in the prompt panel. You can: * Reduce color count to simplify the image and improve print-readiness. * Recolor using a saved palette by clicking on one from your library. There are two modes for adjusting colors, accessible via labeled tabs within the Adjust color panel: * **Swatches** (default): Adjust individual colors by clicking and replacing them. * **Spectrum**: Adjust groups of similar colors all at once. Ideal for complex illustrations with subtle tonal variations. Tip: Spectrum mode is more efficient for batch changes or when fine-tuning a unified look across multiple similar hues. # Using color palettes in generation Source: https://www.recraft.ai/docs/recraft-studio/color-palettes/using-color-palettes-in-generation When generating images, you can apply a color palette to guide the output: * On the Free plan, Recraft offers a rotating selection of preset palettes. These palettes are hand-picked and shuffled for variety and quick inspiration. * On a paid plan, you can create and save custom palettes, which can be applied to any image generation. Palettes can be used with any image type: photorealistic, illustrated, vector, or icon. You can also apply a background color separately to control the image’s overall tone. To apply a palette, click the **Color Palette icon** in the prompt panel. You can also explore randomized inspiration palettes from the “More” menu, useful for experimenting with color direction. Tip: The more complex your prompt or style, the harder it may be for the model to strictly adhere to limited palettes. Simpler prompts and styles tend to yield more accurate color matching. # How to export an image Source: https://www.recraft.ai/docs/recraft-studio/exporting-and-sharing/how-to-export-an-image 1. Select the image or images on the canvas. 2. Click the **Export** button in context panel. 3. Specify your image dimensions, format (PNG, JPG, TIFF, PDF SVG, or Lottie) and resolution. 4. Click the **Export** button. You can also right-click the image and choose **Export as...** from the context menu. Vector images can be exported as SVG and PDF for scalable use. # How to share a project Source: https://www.recraft.ai/docs/recraft-studio/exporting-and-sharing/how-to-share-a-project ## Overview Sharing a project in Recraft allows you to give other users access to your work. This is useful for team collaboration, reviewing work-in-progress, or handing off assets to a client or colleague. ## How to Share a Project 1. From the **Projects** view, hover over the project you want to share 2. Click the **⋯ (three-dot menu)** in the top-right corner of the project card 3. Select **Share project** from the dropdown menu 4. In the sharing dialog, enter the email address of the person you want to invite 5. Set their permission level using the dropdown next to the email field (default is **Viewer**) 6. Click **Invite** to send the invitation To share via link, click **Copy link** at the bottom of the dialog and send it directly. You can control what link recipients can do using the **Anyone with the link** permission dropdown. ## Viewing prompts and exporting assets In a project, you can view the prompts and generation settings in the right-side panel. This allows you to review how assets were created and replicate or adjust results if needed. You can also export images directly from the project for external use. ### Permissions & Access | Role | What they can do | | :------------ | :---------------------------------------- | | **Viewer** | Browse and view all assets in the project | | **Commenter** | View assets and leave comments | Link sharing can be set to: | Setting | What it means | | :------------ | :----------------------------------------------------------------------- | | **No access** | The link does not grant access - only invited users can open the project | | **Viewer** | Anyone with the link can view the project | | **Commenter** | Anyone with the link can view and comment | **Limitations:** * There is no Editor role for shared users — only the project owner can generate or edit assets * Shared users can only access the specific project they were invited to, not the full workspace * Sharing permissions can be updated or revoked at any time by the project owner ### Best Practices **Use sharing when:** * Collecting feedback or review from a client or stakeholder * Giving a teammate visibility into an active project * Sharing a read-only link for quick review without requiring sign-up **Tips for smooth collaboration:** * Use **No access** for the link setting by default, and invite specific people by email for more controlled access * Switch to **Commenter** link access when you need quick feedback from multiple reviewers without managing individual invites * Name your projects clearly before sharing - the project name appears in the sharing dialog and in the recipient's view # How to share an image Source: https://www.recraft.ai/docs/recraft-studio/exporting-and-sharing/how-to-share-an-image You can share an image in multiple ways: * right-clicking on the in the canvas and selecting **Copy image link** * opening the image in the Community, Favorites, or History panels and clicking on the **link icon** to copy the URL. You can then share the URL. A Recraft account is required to view shared images. If the image is public, the link will be viewable by others. On paid plans, you can also set privacy settings before sharing. For **private images** (images generated with a subscription), this link will not open for other users. To share a private image together with its generation settings, you need to create a **project** and add the image there. This allows you to share both the image and its full generation context in one place. # Upscaling Source: https://www.recraft.ai/docs/recraft-studio/format-conversions-and-scaling/upscaling Recraft Studio offers two upscaling options, Crisp and Creative, depending on your needs for resolution and detail. * **Crisp upscale** increases resolution and sharpness without altering the structure or content of the image. Ideal for icons, vector-style illustrations, and assets where visual consistency is critical. * **Creative upscale** enhances resolution while subtly regenerating image content. It can improve anatomy, textures, and fine details, making it well suited for photos, portraits, or group scenes. Note: all creative upscale results are slightly different, and some iterations might render better than others. Note: Creative Upscale is a more resource-intensive process than Crisp Upscale. If you only need to increase resolution, use Crisp Upscale for faster results. If you want to enhance small details, improve anatomy (such as faces or hands), or add definition to background elements, Creative Upscale is the more suitable option. **How to upscale an image**: 1. Select an image on the canvas. 2. In the context panel, click either **Crisp Upscale** or **Creative Upscale**, depending on the image type and desired result. 3. Wait for the upscaled version to appear on the canvas. 4. When you're satisfied, export the upscaled image in your preferred format (e.g., PNG, JPG, PDF). Note: Upscaling doesn't support transparent images. The result of the upscaled image with a transparent background will have a white background instead of a transparent background. # Vectorizing Source: https://www.recraft.ai/docs/recraft-studio/format-conversions-and-scaling/vectorizing You can convert raster images into editable vector graphics directly within Recraft Stuio. Once vectorized, the image becomes scalable without quality loss and fully editable in terms of color. **How to vectorize an image**: 1. Place a raster image (e.g., PNG or JPG) onto the canvas. 2. Select the image, then click the **Vectorize** icon in the context panel. 3. Use the **Color Reduction** slider to limit the number of colors in the output—this helps simplify complex images or prepare for print. 4. Edit colors using two available modes: * **Swatches mode** (default): Click on individual color swatches to replace specific colors. * **Spectrum mode**: Adjust entire ranges of related colors at once—for example, shift all blue tones to green. Ideal for images with many subtle variations. 5. When you’re satisfied, export the vector as an SVG file for use in web, print, or design workflows. **Note**: Not all images can be cleanly vectorized. Photorealistic images, 3D renders, or highly textured illustrations may produce artifacts when converted to SVG, since the format supports only simplified shapes, lines, and fills. # Inpainting Source: https://www.recraft.ai/docs/recraft-studio/frames/inpainting Inpainting lets you regenerate specific parts of an image while keeping the rest intact. This is useful for correcting details, replacing elements, or making creative changes without starting from scratch. Inpainting is currently available in the old interface, which is accessible by clicking the **Old interface** toggle in the top navigation. **How to use inpainting**: 1. Select an image on the canvas. 2. Click the **Edit area** button in the top toolbar. 3. Choose a selection tool: **Lasso**, **Brush**, **Area**, or **Wand**. 4. Highlight the part of the image you want to change. 5. Enter a new prompt describing the desired update. 6. Click the **generation button** to apply the edit. The selected area will be replaced based on your prompt while the rest of the image stays the same. **Note**: When using inpainting on large images, Recraft automatically downscales the image before applying edits, which can result in a noticeable loss of quality in the final output. ## Tips for better results * If the selected area is too small, the change may be subtle or unnoticeable. Select a larger area for more effective inpainting. * Inpainting works best on images resolutions of 1024 pixels or smaller. For larger images, Recraft automatically downscales before editing. The result will be returned at the supported resolution of 1024 pixels. * After multiple rounds of inpainting, image quality may begin to degrade. If that happens, try reverting to an earlier version and batch several edits together in a single step. * To learn more about the selection tools available, visit [Selection tools](/docs/recraft-studio/work-area/selection-tools). # Introduction to frames Source: https://www.recraft.ai/docs/recraft-studio/frames/introduction-to-frames Frames are flexible containers for organising and grouping content on your canvas. Use frames to structure your work, keep related elements together, and share or export specific parts of your canvas with ease. They're especially useful for organising layouts, grouping visuals, and sharing or exporting specific sections of your canvas. **How to use a frame**: 1. Open the toolbar and select **Frame** to add a frame to your canvas. 2. Drag an existing image, text, or shape into the frame, or create new content inside it. 3. Change the frame's background to suit your composition. 4. Move the frame to reposition all its contents together as one group. 5. Rename the frame, export it as a single image, or copy a direct link to share it with others. Frames support images, text, and shapes. Videos, image sets, exploration sets, mockups, and other frames cannot currently be placed inside a frame. > **Note:** Generating a frame, blending all contents, and all associated generation settings are not available in the current UI. # Outpainting Source: https://www.recraft.ai/docs/recraft-studio/frames/outpainting Outpainting expands your image by generating new visual content beyond its original borders. It’s especially useful for changing aspect ratios, extending compositions, or adapting existing images to new formats. ### How to use outpainting (two methods): **Method 1**: 1. Double click on an image on the canvas. 2. A resizable frame will appear around the image. Click and drag the gray bars on any or all sides of the frame to your desired new size. 3. Click the **Expand** button. 4. Optionally, add a prompt in the box below to edit or enhance your image. **Method 2**: 1. Add a **Frame** to your canvas. 2. Drag an existing image into the frame. 3. Leave the prompt unchanged, or add a short prompt if you want to influence the expansion. 4. Click the **generate button** in the prompt panel to generate new content around the original image. 5. Optionally, add additional elements (like text or overlays) inside the frame to include them in the final composition. # History Source: https://www.recraft.ai/docs/recraft-studio/history All images you generate, including images from deleted projects, appear in both the **History page** and the **History panel**. You can access History in two ways: * From the **Recraft home menu** in the top navigation, select **History** to open the full [History page](https://www.recraft.ai/history). * From within a project, click **History** in the top navigation to open the History panel. ### History page The **History page** is the primary place to manage your generated images. It provides full functionality, including: * filtering by type * searching across your entire history * viewing detailed image metadata * copying prompts and image links * downloading images * deleting images (under premium plans) * selecting multiple images for bulk actions Use the History page when you need to review, organize, or clean up your image history. ### History panel The **History panel** opens inside a project and is designed for quick access and reuse. It supports: * filtering by type * searching your image history * viewing detailed image metadata * copying prompts and image links * downloading images * dragging images onto the canvas * inserting a copy of an image into the current project * deleting images (under premium plans) The History panel displays generations in a grid of square tiles. You can filter the content by **Image** or **Video** to quickly find specific generations. The same filtering options are available in the **History panel on Canvas**. Generations are grouped by date, making it easier to browse creations from specific days. ## Browsing and searching (History page) ### Filters You can filter items on the History page using the **Type** filter to display either **Images** or **Videos**. You can also view your history grouped by date for easier navigation. ### Search Use the search bar to find images using simple keywords, such as *red panda* or *rainbow cake*.\ Search matches against the prompt used to generate the image. ## Working with images (History page) ### Hover actions When you hover over an image thumbnail, you can: * Copy the prompt * Download the image * Delete the image (under premium plans) * Select the image for bulk actions (a checkbox appears in the corner) ### Image detail panel Click any image to open its image detail panel. This panel shows: * the prompt used to generate the image * whether the image is raster or vector * the aspect ratio * the model or style used From the image detail panel, you can also: * add the image to Favorites * copy the image URL * download the image * permanently delete the image (under premium plans) ## Bulk actions on images (History page) Bulk selection allows you to perform actions on multiple images at once. ### Selecting multiple images 1. Open the [History](https://www.recraft.ai/history) page. 2. Hover over a thumbnail to reveal the checkbox in the corner. 3. Click the checkbox to select the image. 4. Continue selecting additional images as needed. * A bulk action bar appears at the bottom showing how many images are selected. 5. Click a checkbox again to deselect an image. 6. Click the **X** in the bulk action bar to clear all selections. ### Bulk delete, download, or favorite Once one or more images are selected, the bulk action bar lets you: * **Download** all selected images * **Favorite** all selected images * **Delete** all selected images (confirmation required) **Note**: Bulk delete cannot be undone. You can still regenerate images later using their original prompts. ## Using images in your canvas From the **History panel** inside a project: * Drag an image onto the canvas to place it directly * Or click a thumbnail to insert a copy into your current project Images added from History behave like any other canvas image and can be edited, transformed, or combined with other elements. # Background tools Source: https://www.recraft.ai/docs/recraft-studio/image-editing/background-tools ## Remove background Recraft Studio’s Remove background tool is especially tuned for AI-generated images, making it more effective than the background removal features found in most other AI image generators. You can remove the background from any image with one click, creating a transparent version for compositing or mockups. To remove a background, select the image and click **Remove background** in the context panel. The background will be deleted, leaving only the subject. This is especially useful for preparing product overlays or isolating graphics. Remove background functionality works with images with resolution of 1024 by the smaller side. For images with large resolution there might be some artifacts on the border of the image. **Note**: When using Remove background on large images, Recraft automatically downscales the image before applying edits, which can result in a noticeable loss of quality in the final output. ## Change background **How to change the background of an image**: 1. Select a generated or uploaded image. 2. Select the image, then click **Change background** in the context panel. 3. Wait for the background area to be masked. 4. Enter a new prompt for the background in the prompt panel, then click the **generate button**. 5. Your image with new backgrond will appear on the canvas as a new image. # Editing images with natural language Source: https://www.recraft.ai/docs/recraft-studio/image-editing/editing-images-with-natural-language Recraft Studio supports targeted image editing using integrated external models **Google Nano Banana**, **OpenAI GPT‑4o**, **Seedream**, **Flux Kontext**, and **Qwen-Image** that respond to natural language prompts and optional reference images. These models are built directly into the Recraft interface, so you can make precise edits without switching platforms or starting from scratch. Natural language editing allows you to modify specific aspects of an image (like clothing, background, or facial expression) while preserving the overall structure and style. Only the elements described in your prompt are changed; everything else remains intact. ## Use cases * Make small adjustments, such as changing clothing, background elements, or facial features * Refine visual details without regenerating the entire image * Maintain stylistic or character consistency across multiple images * Add, remove, or modify specific objects within a scene * Adjust visual elements like lighting, color palette, or seasonal context * Update product features or character traits while keeping composition intact * Apply a consistent style or mood using reference images ## How to edit an image with natural language prompts 1. Open a project and upload or generate the image you want to modify. 2. Select the image on the canvas. 3. In the prompt panel, select a model: * **Nano Banana** (normal, Pro, Pro 4K) * **Seedream V4** * **GPT‑4o** (Low, Medium, or High quality) * **Flux Kontext** (Pro or Max quality) * **Qwen-Image** 4. Enter a prompt describing the change (e.g., *“make the jacket red”* or *“change background to nighttime”*). 5. (Optional) Click the **paper clip icon** within the prompt panel to attach reference images: * Up to **9 images** for GPT‑4o, Nano Banana, and Seedream * **1 image** for Flux Kontext and Qwen-Image Reference images allow the model to incorporate visual elements like style, character, pose, or layout into the output. In your prompt, describe what you want to borrow from each image and how you'd like them composed. 6. Click the **generate button** to apply the edit. 7. You can repeat the process as needed. Each new edit builds on the previous result without requiring a full regeneration. ## Supported models Currently, the reference image feature is supported only by select external models. You can manually choose the model, or GPT‑4o Medium will be used by default when a reference image is attached. Recraft standard image generation is included in the table below for reference. | Model | Credits | | ------------------------------- | ---------- | | Recraft raster image generation | 1 credits | | Recraft vector image generation | 2 credits | | Flux Kontext Pro | 2 credits | | Flux Kontext Max | 4 credits | | GPT-4o Low | 1 credit | | GPT-4o Medium | 6 credits | | GPT-4o High | 24 credits | | Nano Banana | 2 credits | | Nano Banana Pro | 15 credits | | Nano Banana Pro 4K | 30 credits | | Seedream V4 | 2 credits | | Qwen-Image | 2 credits | ## Notes on external models Each external model has its own supported aspect ratios. For example, GPT‑4o supports only 1:1, 3:2, and 2:3 formats. If your original image uses an unsupported aspect ratio (e.g., 16:9), the output will default to the closest supported format. The credit cost for using external models depends on each model's API pricing. # How to crop an image Source: https://www.recraft.ai/docs/recraft-studio/image-editing/how-to-crop-an-image To crop an image, hold ⌘ Cmd (Mac) or Ctrl (other systems) and drag its border to your desired dimensions. # Manual editing tools Source: https://www.recraft.ai/docs/recraft-studio/image-editing/manual-editing-tools **Manual selection tools** Recraft Studio provides several selection and modification tools for precise manual edits, such as the **Lasso**, **Brush**, **Area selector**, **Magic wand**, **Eraser**, and **Cut** tools. These allow fine control over what parts of an image are affected when applying edits, regenerations, or style changes. For detailed descriptions and usage of each tool, go to [Selection tools](/docs/recraft-studio/work-area/selection-tools). ## Edit area You can manually select and edit an image or design using the Edit area feature. To activate it, select an image on the canvas, then click **Edit area** in the context panel that appears at the right of the canvas. This will open a panel where you can choose the Lasso, Brush, Area, and Wand tools. To exit these selection tools, click **Edit area** again. ## Duplicate an image There are multiple ways to duplicate an image: * Use the copy-paste function: Cmd/Ctrl+C and Cmd/Ctrl+V * Hold Option on Mac and Alt on other OS, then click and drag the new image * Right-click the image and select **Duplicate** # Raster effects Source: https://www.recraft.ai/docs/recraft-studio/image-editing/raster-effects Recraft Studio effects let you add visual treatments to any **raster image** on the canvas, such as illustrations, photos, or painted textures. These effects do not apply to vector images. You can combine multiple effects, reorder them, or remove them at any time. Effects are non-destructive; you can remove them without modifying the original image. ## Accessing raster effects Raster effects are accessed through the **Adjust colors** control in the context panel. To open raster effects: 1. Select a **raster image** on the canvas. 2. In the right sidebar (context panel), click **Adjust colors**. 3. Click **+** in the **Effects** section to add an effect, such as **Blur**, **Drop Shadow**, or **Stroke**. The Effects section is available only when a raster image is selected. ## Drop shadow Drop shadow adds depth by placing a soft shadow behind your image. To add a drop shadow: 1. Select a raster image on the canvas. 2. In the right sidebar, open the **Effects** section and choose **Drop Shadow**. 3. Adjust the available controls: * **Position (X, Y):** Offsets the shadow horizontally and vertically * **Blur:** Controls how soft or sharp the shadow edge appears * **Color:** Sets the shadow color (use darker tones for natural shadows) Tip: Drop shadow also works with images with transparent backgrounds. ## Stroke Stroke adds an outline around the frame of your raster image. To add a stroke: 1. Select a raster image. 2. In **Effects**, choose **Stroke**. 3. Customize your outline using: * **Weight:** Controls the thickness of the outline * **Color:** Sets the outline color * **Position:** * **Inside** – stroke sits just inside the image boundary * **Outside** – stroke sits outside the image boundary * **Center** – stroke is placed half inside, half outside the boundary ## Blur Blur softens the entire image for atmospheric or stylistic effects. To apply a blur: 1. Select a raster image. 2. In **Effects**, choose **Blur**. 3. Adjust the **Strength** slider to increase or decrease the blur intensity. Blur is frequently used to create depth, soften textures, or reduce detail behind foreground elements. ## Removing an effect Each effect added appears as an entry under **Effects** in the right sidebar. Click **–** next to any effect to remove it. Removing an effect immediately resets all of its settings to default. # Remix Source: https://www.recraft.ai/docs/recraft-studio/image-editing/remix Remix generates new versions of an image while keeping the overall concept intact. Each remix introduces slight differences in elements like character appearance, object position, or scene composition. Remix the image by adjusting the description and similarity strength. Start with an image, rewrite the prompt (e.g., 'cat' to 'dog'), and set how closely the new image should match the original. The Remix function does not use a prompt. Instead, it uses the visual content of the original image to generate alternatives in different aspect ratios. **How to use Remix**: 1. Select an existing image on the canvas. 2. Click on the **Remix image** in the context panel. Image 3. **Adjust similarity to the original**
Use the slider at the top to control how closely the new image should match the original. **Select the model**
Under the *Model* section, choose the generation model. * Here, **V4** is selected. **Choose a style**
In the *Style* section, you can apply a predefined style to the image. * Currently, **No styles** is selected. Image 4. Click **Generate** to produce a set of new outputs based on the selected image. # Vector Editor Source: https://www.recraft.ai/docs/recraft-studio/image-editing/vector-editor The Vector Editor lets you edit every point, curve, and color of a vector image directly in Recraft Studio. It works with any vector on the canvas—both images generated in Recraft and SVG files imported from your computer. There is no need to switch to another tool or add extra export steps. Refine your vector right on the canvas, then export it as an SVG or PDF when you're ready. ## How to open the Vector Editor 1. Double-click any vector image on the canvas. The vector is temporarily split into its individual shapes, and you can click outside to exit. 2. To break a vector into shapes permanently, select it and click "**Convert to frame**" in the panel. Either way, each shape becomes its own editable object. The vector splits into its individual shapes inside a frame, and each shape becomes editable. ## Working with shapes Once a vector is converted to a frame, you can edit each shape independently: * **Move, scale, and rotate** any shape. * **Reorder shapes** with **Send to back** and **Bring to front** to control how they overlap. * **Change colors**: select a shape and pick a new fill color. * **Select multiple shapes** inside a frame by holding Cmd/Ctrl and dragging across them. ## Editing points and curves Double-click a shape to edit its individual points, tangents, and edges: * **Move points and edges** by dragging them. * **Add new points** to a path. * **Add or remove tangents**: hold Cmd/Ctrl and click a point to add or delete its tangents. Drag the tangent ends to adjust the curve. * **Delete points and edges**: select them and press Delete. * **Select multiple points** by dragging across them. Hold Shift and click to add or remove individual points from the selection. * **Transform a selection**: moving, rotating, and scaling also work on multiple selected points. To resize relative to the center of the selection, hold Option on Mac or Alt on other OS. Shapes and points snap into place as you move them, making it easier to keep your design aligned. ## Converting back and exporting When you're done editing, you can: * **If you double-clicked the vector**, simply click outside it when you're done. Your edits are kept, and the vector stays a single image. * **If you used Convert to frame**, you can convert the frame back to a vector image at any point to continue working with it as a single object. **Export the frame as a vector**: if the frame contains only vector elements, you can export it as an SVG or PDF. See [How to export an image](/docs/recraft-studio/exporting-and-sharing/how-to-export-an-image). **Note**: A frame can only be exported as a vector if every element inside it is a vector. If the frame contains raster elements, vectorize them first—see [Vectorizing](/docs/recraft-studio/format-conversions-and-scaling/vectorizing). # Agentic mode Source: https://www.recraft.ai/docs/recraft-studio/image-generation/agentic-mode Recraft Studio's Agentic mode enables users to talk and generate images in Recraft Studio like a design assistant. To activate Agentic mode, select **Agentic** in the generation controls in the prompt panel. ## What's possible with Agentic mode **Session-driven generation & iteration**: Start by describing an idea or intent, and iterate on the visuals using the entire session context. **Brainstorming prompts and design ideas.** Use the Agentic mode interface to help craft and refine prompts or explore different design directions without complex instructions. **Step-by-step workflows**: Multi-step tasks are guided automatically to help move from brainstorming to finished assets in fewer steps. **Smart tips in context**: Get suggestions for next actions, such as upscaling images to higher clarity and resolution, discovering different styles, or improving results within a workflow. **Canvas integration**: Result appears side-by-side on the canvas, preserving context to help compare or build on earlier outputs without resetting. ## How credits work in Agentic mode Agentic mode uses the same credit system as the rest of Recraft Studio. Each interaction with the language model consumes credits, in addition to any credits used for image generation or editing. * Simple, one-off requests may use as little as 1 credit. * Longer conversations with multiple iterations and accumulated context consume more credits. * Agentic credits are separate from image generation and editing costs. For better control over usage, start a new session when switching tasks instead of continuing a long conversation. Credit usage scales with compute: * Quick requests stay lightweight. * Complex, multi-step workflows use more credits. This keeps everyday tasks efficient while accurately reflecting the cost of more advanced sessions. ## Example Agentic mode workflows * Start by asking for help with a prompt for a design task, e.g.: *“I need a hero illustration for a SaaS landing page about teamwork. Help me craft 3 prompt ideas”* * Select one of the suggested prompts, edit it as needed, and try generating in different styles by typing *“generate it in flat style”* or *“generate it in photorealistic style”.* * To change specific details on an image, type in session e.g., *“take 2nd image generated in flat style, but add a dog showing a pet-friendly office”* * Experiment with styles by attaching a reference image, then ask in the session to apply the style to another image * When generating or iterating on images, use a session to specify the aspect ratio, number of images, or to apply Recraft features such remove/replace background, vectorize, create variations, or upscale. * Try using some more advanced design workflows or chain of generations and editing actions e.g.,: * *“Generate a photoreal product shot of a ceramic mug on a plain light background. Then remove the background. Then produce a clean vector interpretation of the silhouette.”* * Use key visuals as image reference(s), then ask to create a marketing assets (such as a landing hero, social cover, email hero), specifying desirable formats, colors or other important details. # Artistic levels Source: https://www.recraft.ai/docs/recraft-studio/image-generation/artistic-levels Artistic levels allow you control the complexity, composition, and visual tone of generated images. It is available in the Recraft V3 model. Settings range from **Simple** to **Eccentric**, offering six levels of control: * Simple levels tend to be clean, minimal, and straightforward, similar to stock-style visuals. * Eccentric levels become more dynamic and expressive, with varied layouts, richer compositions, and creative flourishes. * The levels in between offer increasing degrees of visual complexity, allowing you to calibrate the output to suit your project’s tone and context. **How to use the artistic level slider** 1. Start a new image by entering a prompt. 2. Locate **Artistic level** by clicking on the settings icon in the prompt panel. 3. Click the drop down menu and select your desired setting: Simple, Clear, Artistic, Expressive, Rich, Eccentric. 4. Click the **generate button** to generate your image using the selected visual tone. Tip: Set the Artistic level to **minimum** if you want people to look directly at the camera. # Aspect ratios Source: https://www.recraft.ai/docs/recraft-studio/image-generation/aspect-ratios Recraft Studio supports several preset aspect ratios that can be applied when generating new images. Use the **Aspect ratio selector** located in the prompt panel to choose the desired format. Aspect ratio affects how content is framed and composed, and is especially important when preparing assets for platforms with fixed dimensions, such as social media, thumbnails, or print layouts. **How to generate an image using a custom or unsupported aspect ratio** 1. Generate an image using the aspect ratio closest to your desired one. 2. Create an empty Frame and manually type in the **W** and **H** values to the exact dimensions you need. 3. Drag and drop the generated image into the Frame. 4. Click **Generate frame** to outpaint the empty space and fill the frame to your target size. ### Using the Aspect slider to crop an image You can crop an existing image by selecting it and using the Aspect slider to crop it to your desired dimensions. # Context panel Source: https://www.recraft.ai/docs/recraft-studio/image-generation/context-panel The context panel surfaces actions and properties for the current selection on the [canvas](/docs/recraft-studio/work-area/canvas). It changes dynamically based on what is selected and provides a focused place to inspect, modify, and act on one or more objects. The context panel is selection-driven. It appears only when at least one object is selected on the canvas and updates automatically as the selection changes. ## What the context panel is responsible for The context panel handles object-specific and batch actions, including: * displaying properties for the current selection, such as dimensions and metadata * exposing actions that apply to the selected object or objects * enabling batch operations when multiple objects are selected * providing access to export and arrangement controls It does not create new content or interpret natural language instructions. Those actions are handled by the [prompt panel](/docs/recraft-studio/image-generation/prompt-panel). ## Context panel states The context panel adapts based on selection state and object type. Available actions vary depending on whether a single object or multiple objects are selected, and on the types of objects in the selection. **Single object selected** When a single image or object is selected, the context panel shows properties and actions specific to that object. This may include: * object dimensions (width and height in pixels) * prompt and generation-related metadata (when applicable) * object-specific actions such as export, background removal, vectorization, or upscaling This state is used for inspecting and acting on individual elements. **Multiple objects selected** When multiple objects are selected, the context panel switches to batch actions. The set of available actions depends on the types of objects selected. For standard images, batch actions may include: * running the same prompts across all selected images * background removal * vectorization * crisp or creative upscaling * merging layers * exporting selected images For other object types, such as mockups, fewer batch actions may be available. In these cases, the context panel typically includes: * merge layers * arrange * export In all multi-selection cases, an **Arrange** control is available. This control allows you to group and rearrange selected objects along a horizontal or vertical plane. ### Actions for a single selected object When a single image or object is selected, the context panel may include the following actions: * **Reuse in a new image**: Use the selected image’s prompt and generation settings as a starting point for creating a new image. * **Generate Remix**: Create multiple variations of the selected image while preserving its overall structure. View [Remix](/docs/recraft-studio/image-editing/remix). * **Remove background**: Automatically remove the background from the selected image. View [Remove background](/docs/recraft-studio/image-editing/background-tools#remove-background). * **Change background**: Replace the background using natural language instructions. View [Change background](/docs/recraft-studio/image-editing/background-tools#change-background). * **Edit area**: Apply natural language editing to a specific region of the image. View [Edit area](/docs/recraft-studio/image-editing/manual-editing-tools#edit-area). * **Extract prompt**: Retrieve the prompt and settings used to generate the image for reuse or refinement. View [Extract prompt](/docs/recraft-studio/image-generation/working-with-text-and-prompts/extract-prompt). * **Use in mockup**: Place the selected image into a mockup for product or scene visualization. View [Mockups](/docs/recraft-studio/mockups/how-to-create-a-mockup). * **Vectorize**: Convert the selected raster image into an editable vector format. View [Vectorizing](/docs/recraft-studio/format-conversions-and-scaling/vectorizing). * **Crisp upscale**: Increase image resolution while preserving structure and detail. View [Upscaling](/docs/recraft-studio/image-editing/format-conversions-and-scaling/upscaling). * **Creative upscale**: Increase image resolution while enhancing details and textures. View [Upscaling](/docs/recraft-studio/format-conversions-and-scaling/upscaling). * **Adjust colors**: Modify the color palette of the selected image. View [Color palettes](/docs/recraft-studio/color-palettes/using-color-palettes-in-generation). * **Export**: Export the selected object in supported formats. View [How to export an image](/docs/recraft-studio/exporting-and-sharing/how-to-export-an-image). ### Actions for multiple selected objects When multiple objects are selected, the context panel provides batch actions, including: * **Align**: Align selected objects to an edge (left, right, top, bottom, center). * **Arrange**: Adjust the relative ordering of selected objects as a group. * **Merge objects**: Combine selected objects into a single image. * **Export**: Export all selected objects together. View [How to export an image](/docs/recraft-studio/exporting-and-sharing/how-to-export-an-image). # Exploration Mode Source: https://www.recraft.ai/docs/recraft-studio/image-generation/exploration-mode Explore first, refine later: a new way to generate images When people first start working with generative image models, they often assume that the key to good results is writing the perfect prompt. It’s easy to fall into the habit of carefully constructing long descriptions, adding more keywords, and tweaking the text again and again in search of the right output. In practice, creative work rarely begins with precision. Most design processes start with exploration — trying different directions and gradually refining the ideas that feel promising. The process is visual, iterative, and often unpredictable. Exploration Mode in Recraft V4 is designed around that reality. Instead of requiring a perfectly engineered prompt before anything can happen, it allows you to start with a simple idea and explore visual directions first. ## **Starting With a Simple Prompt** In Exploration Mode, the process begins with something intentionally minimal. Rather than describing every detail of an image, you can start with a short prompt that simply defines a direction.\ For example: * Typography poster * Playful 3D mascot * Minimal serif logo 69c3ebf7126d92d78c2ca1ca 1 Once you click **Generate**, the system produces eight different visual interpretations of the prompt. Each result explores a different combination of composition, structure, lighting, and visual style.\ At this stage, the goal is not to produce a finished design. Instead, the system provides a quick overview of possible visual directions so that you can see which ideas feel worth developing further. ## **Seeing Multiple Directions at Once‍** Consider a simple prompt such as **“Stylized illustration of a dog on a clean single-color background.”**\ Rather than generating a single result, Exploration Mode produces eight distinct variations that explore different ways the concept might be expressed. Exploration mode gives V4 more freedom in interpretation. This allows the model to be more creative and give unique and diverse results that may otherwise be impossible to create with regular prompting. This sometimes comes at the cost of slightly less consistent anatomy or prompt adherence at first steps. 69c3f0f9d08a95f95bb20405 2 2 2 ## **‍From Exploration to Refinement‍** Once the initial set of images appears, the next step is to select the result that feels closest to your intention. At this point the workflow naturally shifts from exploration to refinement. Instead of rewriting prompts or starting again, you can continue building on the image that already captures the core of the idea.\ From here, the system allows you to generate new variations based on that selected image.**‍** ## **Refining With Similarity Levels‍** To make this process more controllable, Exploration Mode introduces five similarity levels. These levels determine how closely new results should follow the image you selected. Each level provides a different balance between creative exploration and precision, allowing you to refine ideas gradually without losing the visual direction you discovered.**‍** * **A little bit similar** — this level maintains only a loose connection to the original image. It is useful when you want to push the concept in a new direction while still keeping a hint of the initial idea.**‍** * **Moderately similar** — at this level the overall subject or composition remains recognizable, but stylistic changes become more noticeable. * ‍**Quite similar** — this level maintains strong visual continuity. The composition, subject, and overall mood remain consistent, while smaller refinements appear in lighting, styling, or structure.**‍**‍ * ‍**Very similar** — here the results stay very close to the original image. Only minor changes occur in texture, detail, or micro-composition.**‍** * **Extremely similar** — this level produces near-duplicate variations and is best used for final refinements such as adjusting subtle shape details, reflections, or alignment. Instead of rewriting the prompt repeatedly to force a specific outcome, you simply select the direction that feels most promising and continue developing it. This shift— from describing ideas in text to reacting to visual results — makes the creative process feel both faster and more intuitive. 69c3f13acc1b6aaafdece735 2 ## **Where Exploration Mode Works Best** Exploration Mode is particularly useful in situations where the goal is not to produce a final image immediately, but to explore visual possibilities quickly. * **Concept development** — quickly testing multiple visual directions before committing to one. * **Brand exploration** — generating variations of mascots, visual styles, or graphic motifs. * **Poster and layout ideas** — experimenting with composition and typography before refining the design. * **Campaign visuals** — exploring stylistic directions for marketing assets. * **Creative discovery** — finding unexpected visual approaches that might not emerge from a carefully written prompt. In these contexts, the ability to generate multiple interpretations of a simple idea significantly accelerates the early creative phase.**‍** ## **Designing Through Exploration‍** With exploration mode there is no need to perfectly describe the final image from the beginning, it supports a workflow that feels closer to real creative practice: exploring directions, reacting to visual ideas, and gradually refining what works. Prompts remain simple, images evolve visually, and design decisions emerge through iteration. Most ideas do not start fully formed. They develop through exploration. Exploration Mode is built to support exactly that process. 69c3ec755c35cbc59a7f3f3f 4 # External models Source: https://www.recraft.ai/docs/recraft-studio/image-generation/external-models External models let you utilize top AI models directly in Recraft Studio, so you can generate and edit images without switching tools. | Model | Creator | Description | Time | Credits | | ---------------------- | ----------------- | --------------------------------------------------------------------------------------------------------- | ---- | ------- | | Flux 2 Max | Black Forest Labs | Maximum consistency for complex, multi-image workflows | 20s | 18 | | Flux 2 Dev | Black Forest Labs | Open-weight model for fast experimentation and pipelines | 6s | 3 | | Flux 2 Pro | Black Forest Labs | Production-grade realism with strong prompt alignment | 10s | 10 | | Flux 2 Flex | Black Forest Labs | Configurable quality, speed, and typography control | 20s | 13 | | Flux 1 Dev | Black Forest Labs | Developer-focused version for experimentation | 9s | 7 | | Flux 1.1 Pro | Black Forest Labs | Refined Flux model with high-fidelity rendering | 6s | 8 | | Flux 1 Kontext Pro\* | Black Forest Labs | Context-aware Flux variant preserving semantic alignment in scenes | 7s | 8 | | Flux 1 Kontext Max\* | Black Forest Labs | Advanced Kontext version with enhanced resolution and cinematic detail | 12s | 15 | | GTP Image 1.5 Low\* | OpenAI | Fast, lightweight layout-aware image generation | 15s | 2 | | GTP Image 1.5 Medium\* | OpenAI | Balanced quality and consistency for general use | 25s | 5 | | GTP Image 1.5 High\* | OpenAI | High-detail images with precise layout control | 50s | 24 | | GPT Image 1 Low\* | OpenAI | Lightweight GPT-4o variant for text-to-image tasks, multi image editing | 15s | 2 | | GPT Image 1 Medium\* | OpenAI | Balanced GPT-4o variant with improved detail and realism, multi image editing | 35s | 8 | | GPT Image High\* | OpenAI | High-quality GPT-4o setting delivering photorealistic outputs, multi image editing | 60s | 30 | | HiDream I1 Dev | HiDream AI | Experimental diffusion model optimized for artistic compositions | 12s | 6 | | Ideogram V3 Turbo | Ideogram AI | Fast inference mode optimized for speed | 9s | 6 | | Ideogram V3 Default | Ideogram AI | Standard setting balancing speed and quality | 12s | 11 | | Ideogram V3 Quality | Ideogram AI | Highest-fidelity Ideogram variant | 12s | 17 | | Imagen 4 | Google DeepMind | Advanced Imagen with higher realism | 8s | 8 | | Imagen 4 Ultra | Google DeepMind | Top-tier Imagen for ultra-photorealistic rendering | 10s | 13 | | Nano Banana\* | Google Gemini | Standard-resolution outputs, reliable generation, and foundational editing. | 12s | 7 | | Nano Banana Pro\* | Google Gemini | Faster generation, higher fidelity, accurate multilingual text, and advanced creative controls | 15s | 32 | | Nano Banana Pro 4K\* | Google Gemini | High-resolution variant delivering sharper detail and 4K-grade image outputs. | 30s | 63 | | Seedream V4.5\* | ByteDance | Structured, reference-driven high-fidelity image generation | 30s | 8 | | Seedream V4\* | ByteDance | Next-gen diffusion model delivering vivid colors, fine detail, and artistic realism | 20s | 7 | | Qwen-Image\* | Alibaba | Multimodal model built for high-fidelity image generation and text–image alignment | 7s | 4 | | Z-Image Turbo | Alibaba | Ultra-fast photorealism with strong bilingual (English/Chinese) text rendering and solid prompt adherence | 3 | 1 | \*Used for [editing images with natural language](https://www.recraft.ai/docs/recraft-studio/image-editing/editing-images-with-natural-language) within Recraft. ## **FAQ** **Why do different models have different credit prices?** Different models have different credit prices because each model has unique capabilities and API costs. More advanced or higher-quality models require more credits per use. For example, Nano Banana Pro 4K costs more credits than basic raster generation. The credit cost reflects the model’s complexity and external API pricing. You can find more details about how credits work [here](https://www.recraft.ai/docs/plans-and-billing/credits). # Generating multiple images Source: https://www.recraft.ai/docs/recraft-studio/image-generation/generating-multiple-images ## Images per generation You can generate between 1 and 4 images per prompt by wihthin the image number selector in the prompt panel. This is useful for exploring multiple variations at once without needing to re-enter your prompt. The selected number of images will appear side by side on the canvas after generation. ## Image sets Image sets enable you to generate six visuals at once, each from a different prompt, while keeping them stylistically consistent. Use image sets to generate images with multiple prompts at once using shared settings like style, palette, aspect ratio, etc. This feature is particularly useful when working with icons, illustrations, or batch assets for UI, marketing, or product design. **How to create an image set**: 1. Right-click on the **Canvas**, then choose **Image set** from the toolbar. 2. Enter up to **six individual prompts**, each describing one image (e.g., "Sun icon", "Moon icon", etc.). 3. Select a style from the curated library or your saved/custom styles. You can also adjust the image settings (avoid text and negative prompt) colors. 4. Set the artistic level and aspect ratio as needed. 5. Click **Generate** in the Image set prompt panel to generate the full set. Once the set is generated, you can refine individual results by clicking on any image in the set and updating its prompt or parameters. This regenerates only the selected image while leaving the rest of the set unchanged. # Generation modes Source: https://www.recraft.ai/docs/recraft-studio/image-generation/generation-modes Recraft Studio offers three ways to generate images: **Agentic**, **Manual,** and **Exploration mode**. Each mode is designed for a different workflow, and you can switch between them at any time. Agentic and Manual modes are available in the new interface. If the **Old interface** toggle is enabled, these modes are not accessible. ## AI chat AI chat uses an LLM-powered assistant to help generate and refine images through a conversational workflow. In this mode: * The system understands context and intent across multiple turns, and keeps track of your history. * You can describe ideas naturally, refine them, and ask for suggestions. * The assistant may suggest adjustments, apply styles, or select models automatically when helpful. * You don’t need to restate details repeatedly as you iterate. * Image generation and edits follow the flow of the conversation. Agentic mode is best for exploration, iteration, and situations where you want guidance or are still shaping what you want to create. It may also be faster in *achieving your goal*, especially when you’re not sure how to prompt something. ## Manual (Create) mode Manual mode is a manual, prompt-based workflow with no conversational context. In this mode: * Each request is treated independently. * No history, context, or memory is applied between generations. * You control all inputs explicitly through the prompt and settings. * Generation is faster and more predictable. Manual mode is ideal when you know exactly what you want, need precise control, or are making quick, one-off generations. ## Exploration Mode [Exploration mode](https://www.recraft.ai/docs/recraft-studio/image-generation/exploration-mode) helps you quickly generate and browse multiple visual directions from a single idea. In this mode: * You start with a prompt and receive a variety of generated options at once. * The system focuses on diversity, showing different styles, compositions, and interpretations. * You can easily compare results and pick the direction you like most. * Selected outputs can be refined further or used as a starting point for iteration. * It’s designed to surface ideas you might not have considered. Exploration mode is best when you want inspiration, are comparing creative directions, or need to quickly discover what works before refining further. ## Switching between modes You can switch modes directly in the prompt panel: 1. Locate the mode selector in the prompt panel. 2. Select **Agentic** for conversational, context-aware generation. 3. Select **Manual** for manual, stateless prompt-based generation. 4. Select **Exploration** to generate multiple variations from a single idea and quickly compare different directions. Switching modes does not reset your project or canvas: * Existing images remain unchanged. * You can explore ideas in Agentic mode, then switch to Manual mode to fine-tune details. * You can move between modes freely as your workflow changes. ### Return to the old interface (optional) If you re-enable the **Old interface** toggle, the new interface is disabled and Agentic mode is no longer available. ## When to use each mode | Mode | Best for | Behavior | | --------------- | ---------------------------------- | ------------------------------------------------------------- | | **Agentic** | Exploration, iteration, guidance | Context-aware, conversational, may suggest or adjust settings | | **Manual** | Precision, speed, manual control | Stateless, prompt-only, fully explicit | | **Exploration** | Inspiration, comparison, discovery | Multiple variations, diverse outputs | All modes work within the same canvas and support Recraft Studio’s full set of image generation and editing tools. # How to generate a logo Source: https://www.recraft.ai/docs/recraft-studio/image-generation/how-to-generate-a-logo 1. **Create a new image**. Open your project and add a new image frame on the canvas. 2. **Select a vector style**. For the best results, choose one of Recraft’s vector models and styles designed for logo creation, such as Geometric Logo, Sharp Drawn Logo, or Playful Typographic. You can find more by typing "logo" in the style search bar. 3. **Write your prompt**. Describe the logo you want to create, for example: “logo ancient Greek sculpture, clever negative space”. To include text in the logo, place it in quotation marks. For instance, typing “Broccoli” will add that word as stylized text. 4. **Click the generate button**. Your logo will be generated in seconds. # How to generate an image Source: https://www.recraft.ai/docs/recraft-studio/image-generation/how-to-generate-an-image 1. Create a new project, or go to [http://recraft.new](http://recraft.new). 2. Describe your image in the prompt panel at the bottom of the canvas, then click the **generate button**. After a few seconds, the new image will appear on the canvas in a bounding box. ### Image generation options Use the buttons in the prompt panel to adjust how your image is generated: * **Agentic** / **Manual**: Select between [Agentic and Manual modes](generation-modes). * **Model**: Choose which image model to use. Different models are optimized for different output types, such as raster images or vector graphics. * **Style**: Browse curated styles or explore an infinite feed of AI-generated styles. You can apply a predefined style or create your own using reference images. * **Ratio**: Set the aspect ratio of your image to control its dimensions and framing, such as square, landscape, or portrait formats. * **Count**: Choose how many images to generate from a single prompt, up to 4. Multiple images are generated as variations and appear together on the canvas. * **Color**: Apply a predefined or custom color palette to guide image generation. # Prompt panel Source: https://www.recraft.ai/docs/recraft-studio/image-generation/prompt-panel The prompt panel is where image generation and natural language editing begin in Recraft Studio. It provides access to generation modes, models, styles, and settings, and adapts based on whether you are creating new content or editing an existing selection. ## What the prompt panel controls The prompt panel handles all prompt-driven actions, including: * generating new images, frames, mockups, and image sets * editing selected images using natural language * selecting models, styles, and generation modes * configuring generation parameters such as aspect ratio and number of images It does not manage layout, selection, or object-specific actions such as alignment, arrangement, or export. Those are handled on the [canvas](/docs/recraft-studio/work-area/canvas) and in the [context panel](/docs/recraft-studio/image-generation/context-panel). ## Prompt input and controls The prompt panel includes controls for configuring how images are generated or edited: * **Prompt field**: Enter natural language descriptions or instructions * **Generation mode**: Choose how prompts are interpreted (for example, Agentic or Manual) * **Model selection**: Select the model used for generation or editing * **Style selection**: Apply visual styles to generation * **Aspect ratio**: Control image dimensions * **Images per generation**: Generate multiple outputs at once Some controls are available only in specific modes or when editing an existing selection. ## Prompt panel states The prompt panel changes behavior based on selection state. **No selection** When nothing is selected, the prompt panel is in generation mode. You can: * enter a prompt to generate new images * choose a model and style * set aspect ratio and number of images * start new generations This is the default state for creating new content. **Single object selected** When a single image or object is selected, the prompt panel switches to editing mode. You can: * modify the prompt associated with the selected image * apply natural language edits or refinements * regenerate variations based on the current selection This state is used for iterating on existing images rather than creating new ones. **Multiple objects selected** When multiple images are selected, the prompt panel switches to **reference mode**. In this state: * the selected images are treated as **reference images** for a new generation * any prompt you enter is interpreted in relation to the selected images * the next generation creates a new image informed by those references, rather than editing them directly This state is used to generate new content guided by visual examples on the canvas. ## Docking and placement The prompt panel can be toggled between different positions depending on your workflow: * docked at the bottom of the canvas * docked to the side of the workspace using the prompt panel toggle Docking affects layout only. All prompt panel functionality remains the same regardless of position. ## FAQ Can I see the prompts used for images on the canvas? Yes, you can view the prompt and other generation details for any image you’ve created or added to your project. When you select an image on the canvas, the Context panel will show properties including the prompt used to generate it, along with model, style, aspect ratio, and other settings. # Selecting a style Source: https://www.recraft.ai/docs/recraft-studio/image-generation/selecting-a-style By default, images you generate are created by top-ranked Recraft’s V3 model. If you want to generate an image in a different style, click the **Style** button in the prompt panel to open the Style library. You will see two primary style libraries: * The **Curated** tab contains styles categorized by style type: Photorealism, Illustration, Vector art, Vector icon, Graphic design, Vector logo, and Other styles. * The **AI Generated** tab contains an "infinite" collection of AI-generated styles Select and set your style, then click **Generate button** to see the results. For more about styles, go to [Styles overview](/docs/recraft-studio/styles/overview). # Adding text to an image Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/adding-text-to-an-image Recraft Studio lets you add and manipulate text directly on the canvas, either as part of your layout or as input for generation. Text can be layered manually over any image, used within a **Frame** to guide generation of images with text, or incorporated into prompts for image generation. ## Typing text into the prompt In Recraft Studio, you can generate text as part of an image by simply including it in the prompt. This is useful when you want the text to appear stylized within a design, such as a logo, sign, label, or poster, rather than adding it manually afterward. You can write prompts that include both a visual description and the desired text, or prompts that only describe the text itself. **Examples:** * `"A coffee shop logo with the words 'Java Joint'"` * `"The phrase 'Open super late' in neon sign style"` * `"Tote bag design with the words 'Support your local bookworm'"` Recraft Studio’s models will interpret the text prompt and incorporate it directly into the image, styled to match the overall look. You can increase accuracy by placing the text in quotation marks, which helps the model recognize it as literal text rather than a text description. This method is best suited for cases where the text should feel embedded in the visual design rather than overlaid. ## Adding text manually Manual text allows font selection from Google Fonts, color, size, line height, and letter spacing adjustments. This is best used for static overlays (e.g., captions, annotations, headlines) where you want full manual control. How to manually add text: 1. On the canvas, select the **text tool** from the left toolbar. 2. A blank cursor appears. Click anywhere on the canvas and start typing. 3. Use the **text context panel** to adjust typeface, font size, weight, line height, letter spacing, color, and alignment as needed. 4. Use the bounding box controls to resize, rotate, or reposition the text as needed. 5. You can combine text with other visuals, including images, mockups, or frames. ## Using text in a frame for AI layout generation For more advanced compositional control: 1. Add your image to the canvas. 2. Create a **Frame** and place the image and any text elements inside it. 3. Select the frame, then write a prompt (e.g., "Design a layout using the image and title text"). 4. Click **Recraft Frame**. Recraft Studio will use both the visual and text elements as input, generating a new composition with the text integrated according to the chosen style. This is useful for designing social media graphics, posters, or promotional layouts where you want AI assistance with composition and visual harmony. **Notes and limitations** * You can include multiple text elements within a Frame to simulate layout or multi-line designs. * Recraft Studio uses the position of text boxes in a Frame as layout suggestions. During generation, text may be repositioned or restyled based on the selected style. * Text added via a Frame becomes part of the generated image and cannot be edited afterward. * For vector images, any text included will be preserved in the exported SVG file. * Recraft V3 supports generating images with text, but does not handle very small text reliably. If text appears distorted or contains typos, try increasing its size. * Recraft V2 does not support image generation with text. Using this model may result in text rendering errors or misspellings. * Recraft V3 supports text generation within the following supported characters: '!', '"', '#', '\$', '%', '&', '\\'', '(', ')', '\*', '+', ',', '-', '.', '/','0', '1', '2', '3', '4', '5', '6', '7', '8', '9',':', '\<', '>', '?', '@','A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R','S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z','\_', '', 'Ø', 'Đ', 'Ħ', 'Ł', 'Ŋ', 'Ŧ', 'Α', 'Β', 'Ε', 'Ζ', 'Η', 'Ι', 'Κ', 'Μ', 'Ν', 'Ο', 'Ρ', 'Τ','Υ', 'Χ', 'І', 'А', 'В', 'Е', 'К', 'М', 'Н', 'О', 'Р', 'С', 'Т', 'У', 'Х','ß', 'ẞ'.\ Notes: * Characters in different cases (like “B” and “b”) are treated as the same symbol. * Characters that appear identical are different characters with minute variations. * Characters not listed in this character set will not render. # Extract prompt Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/extract-prompt **Extract prompt** lets you generate a descriptive text prompt from any image, whether it was made in Recraft Studio or uploaded from elsewhere. The tool analyzes the visual content and returns a detailed description of what it “sees,” including aspects like subject, composition, technique, and style. This is especially useful when you want to: * Understand how to describe an existing image in prompt form * Recreate or iterate on a similar image * Explore stylistic attributes for reference or learning * Reuse successful results in new contexts * Refine previous generations without starting from scratch * Analyze how a particular outcome or style was achieved Unlike reusing a generation’s original prompt, **Extract prompt** works as an AI interpretation. It does not recover the exact text that was originally used to generate the image. #### How to use Extract prompt 1. Upload an image from your computer or select one on your canvas. 2. Click **Extract prompt** in the context panel. 3. Recraft will analyze the image and display a text description summarizing its content and visual style. 4. You can edit this description or use it directly as a new prompt for generation. *Tip:* Extract prompt works best with clear, well-defined visuals (illustrations, product photos, icons, etc.). The richer the image detail, the more specific and useful the extracted prompt will be. # Negative prompts Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/negative-prompts Negative prompts help guide the AI by specifying what you don’t want to appear in the image. This gives you more control over the outcome, especially when avoiding unwanted objects, styles, or features. **How to use negative prompts**: 1. Select your image on the canvas and click the **settings button** in the prompt panel. 2. In the **Negative prompt** field, enter the elements or concepts you want to exclude from the image. 3. Optionally, you can disable text generation in your output by toggling the **Avoid text in image** switch to **Yes**. 4. Generate your image as usual. Recraft will attempt to avoid the elements you specified. Tip: avoid double negation in negative prompts. For example, to exclude apples from an image, write `"apples"` in the negative prompt, not `"no apples"`. Using “no” can confuse the model and lead to the opposite effect. # Using custom fonts Source: https://www.recraft.ai/docs/recraft-studio/image-generation/working-with-text-and-prompts/using-custom-fonts You can upload your own font files and use them inside Recraft when adding text. Custom fonts are particularly useful for brand typography, logo refinement, packaging mockups, marketing visuals, and consistent presentation design. ## How to upload a custom font 1. Select the **Text** tool. 2. Go to the Context panel on the right, locate **Typeface** and click the **upload icon**. 3. Select your font file (`.woff`, `.woff2`, `.ttf`, or `.otf`). 4. The font will appear in your font list. 5. Apply it to any text layer. ## Important notes * Uploaded fonts are available within your workspace. * If a font file does not include certain character sets (for example, Cyrillic characters), those characters will not render correctly. * Ensure the font file contains all required glyphs for your language and use case. # About importing Source: https://www.recraft.ai/docs/recraft-studio/importing/about-importing Recraft Studio lets you bring your own images into the canvas. Imported images can be used across the platform’s tools, from mockups and vectorization to prompt-based editing and fine-tuning. Once imported, your image can be used to: * Generate a variation of the original visual * Use as a reference for prompt-based generation * Remove the background * Vectorize raster images (e.g., PNG to SVG) * Remix or edit specific areas with prompts * Apply your image to a product mockup * Extend the image using outpainting (frame tool) * Create a custom style for consistent generation * Upscale to increase resolution and detail **Notes on imported and edited images** * Only **original generated images** are saved to project history. Images that have been edited e.g., background removed or fine-tuned) will not appear in the project timeline. * Imported images, whether added to the canvas or uploaded to create a custom style, are not used for training AI models. * To reuse edited images, first **export them**, then **re-import** as needed. * When importing an image, Recraft Studio may default to an external model to enable natural language editing. # How to import an image Source: https://www.recraft.ai/docs/recraft-studio/importing/how-to-import-an-image There are a few ways to import an image: * Drag and drop an image directly from your system file browser onto the canvas * Click the **Upload image** button in the **toolbar** * Click **History** in the top navigation to open the history panel. Click on any image to place it onto the canvas. Once placed, an imported image will behave like any other canvas element. You can resize, rotate, and reposition it as needed. **Note**: Imported images are not used for training AI models. # Size limits and supported formats of imported images Source: https://www.recraft.ai/docs/recraft-studio/importing/size-limits-and-supported-formats JPG, PNG, WebP: * Width and height dimensions should not exceed 5800 pixels each * Max resolution: 16 megapixels * Max file size: 48MB SVG: * Max file size: 48MB # Interface language Source: https://www.recraft.ai/docs/recraft-studio/interface-language Recraft is designed for creators around the world. To make the experience more comfortable and accessible, the interface is now available in **six languages**. You can easily change the interface language directly from your **Profile settings**. ### How to change the language 1. Open your **Profile** in Recraft. 2. Locate the **Language** dropdown menu. 3. Select your preferred language from the list. Once selected, the interface will automatically update. ### Available languages Currently, Recraft supports the following languages: * **English** Image * **Deutsch (German)** Image * **Español (Spanish)** Image * **Français (French)** Image * **Українська (Ukrainian)** Image * **Русский (Russian)** Image You can switch languages anytime from the **Profile → Language** dropdown menu, and the change will apply instantly across the interface. # Interface overview Source: https://www.recraft.ai/docs/recraft-studio/interface-overview This product walkthrough highlights the three available session layouts, explains how Manual, Agentic, and Exploration modes differ, and shows where to configure models, styles, and generation settings. It also demonstrates how generated images connect to on-canvas tools such as background removal, edit area, vectorization, and mockups, and how to bring images to life using the built-in video generation tools.