Generative fill could be even more useful if it could interpret prompts of "same thing but..." to make alternations to what's already in the image instead of replacing it. For example, "same face, but eyes looking left and mouth open," or "same armor but more damage."
AdChoices