A generated face looks warm and natural in one shot, then suddenly gray or orange after a cut. That can make a short video feel stitched together even when the facial structure is consistent.
The cause may be a change in source lighting, uneven reference photography, or the model interpreting each shot with slightly different color balance.
Compare the surroundings first
If the entire scene becomes cooler after a cut, the lighting may actually have changed in the original footage. The face should respond to that change. If only the replaced face changes color while the neck and background stay stable, the edit is less convincing.
Don't demand a constant skin color under every light
Real skin looks different in warm sunlight, blue stage light and a dark indoor scene. The goal is consistent identity under changing illumination, not identical pixel values in every shot.
A useful prompt: “Maintain the reference person's natural skin appearance while adapting realistically to the source scene's lighting. Keep face and neck illumination consistent within each shot.”
Pick a neutral reference
A portrait with heavy color grading can influence the generation in unwanted ways. Use a reference with believable color and enough midtone detail to show the cheeks, forehead and jaw.
When the AI preserves the right features but struggles with color, that is a different diagnosis from wrong identity mapping. Keep those questions separate.
Review the edit on more than one screen
A saturated display can hide color casts or exaggerate contrast. Check the final MP4 in a normal player and, if possible, on a second device. Avoid making major color corrections based on one highly customized monitor setting.
Good color continuity makes an AI edit feel integrated into a scene. It is worth checking, but only after the faces themselves are correctly mapped.
