> ## Documentation Index
> Fetch the complete documentation index at: https://www.recraft.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# 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.