Stop ChatGPT from changing faces.

Before you begin

Upload ONE approved character image that you have permission to use. Name the scene you want next. Start with one image, not a whole video.

Short prompt

Use the attached image as my character reference. Keep the face, skin tone, hair and proportions. Change only [pose/background]. Generate one image for review.

Full workflow prompt

Use the attached image as the visual reference for this character. Preserve the recognizable facial structure, skin tone, age appearance, body proportions, hairstyle, hair color and rendering style.

Create ONE new image for this scene: [describe scene]. Change only: [pose, expression, background or other approved changes]. Keep unchanged: [wardrobe, accessories and other fixed details]. Output shape: [2:3 portrait or other ratio].

If a detail is ambiguous, ask before generating. Do not add a new character, change the face or redesign the outfit. If this tool cannot generate or edit images, return a production brief instead and say so. After generating, identify visible inconsistencies if you can inspect the output; do not claim a perfect match.

Review the result

Compare the eyes, nose, jaw, skin tone, hairstyle and proportions side by side with the approved reference. Also check hands, clothing and accessories. Approve the still before requesting another scene or animation.

Refine it

Return to the original approved reference. The latest version changed [specific feature]. Correct that feature while keeping [approved scene details]. Make one corrected image. Do not use the incorrect face as the new reference.

Put it to work

Example brief: same character, same burgundy scrubs and curly ponytail, seated at a kitchen table reviewing a notebook. Change only the setting and seated pose. This is a practice brief, not a measured result.