Character reference
A character reference helps a compatible Lewdora model interpret recurring visual traits such as face shape, hair, color palette, and overall identity. It is guidance, not an exact-copy guarantee.
Short answer
Character-reference AI uses an uploaded image to guide identity in a new generation. In Lewdora, supported models can combine that reference with a new prompt, pose, scene, and style.
A clear, unobstructed character image usually gives the system a more useful identity signal than a distant or crowded scene. Consistent lighting and visible facial traits make it easier to tell whether a later output preserved the features that matter.
Let the reference carry recurring visual traits, then use the prompt to describe the new expression, clothing, pose, camera, lighting, and environment. Conflicting descriptions can pull the result away from the reference, so keep identity language stable while changing the scene.
Reference controls are in beta and may interpret an image loosely. Compare several outputs using the same prompt and settings before deciding whether a workflow is consistent enough for your project. Do not claim exact identity preservation when the evidence shows variation.
Workflow
Choose a permitted image where the character traits you care about are visible.
Describe recurring traits consistently across prompts.
Adjust pose, clothing, setting, or lighting without rewriting the identity.
Check face, hair, palette, and silhouette rather than judging one image.
Learn how to write stronger prompts, use references, manage privacy, and choose the right pass.
Set subject, style, setting, and limits before spending tokens.
Read guide
Use image inputs to steer character, pose, and structure.
Read guide
Match model strengths to linework, mood, and finish.
Read guide
Keep, download, and delete browser-local images with intent.
Read guide
Pick a pack or 30-day pass based on how often you create.
Read guide