Requested a new background. Received a new relative.
Why AI Image Editing Changed the Person’s Face
Faces contain many tightly related details, and generative editing may reconstruct them when nearby pixels change. Even small differences in eyes, proportions or skin texture can make a familiar person look noticeably different.
Why this happens
- The edit mask overlapped the face or its surrounding features.
- The tool regenerated the whole image at low identity fidelity.
- The prompt described a person instead of preserving the supplied identity.
- Repeated edits compounded small changes.
What to try next
- Keep the face outside the edit selection.
- State that identity, expression, pose and facial geometry are locked.
- Return to the original for each new edit rather than editing an edit.
- Use an identity-preserving or conventional compositing workflow when likeness matters.
Frequently asked questions
Can a prompt perfectly preserve a face?
It can help, but the tool and masking controls matter more than wording alone.
Why do repeated edits make it worse?
Each generation can introduce small identity changes that accumulate over successive versions.