An AI image edit changed the subject it was told to preserve
The request required pixel-level preservation, while a generative editor redrew the whole image region.
What the user wanted
The actual target
Remove the background from a product photo while leaving the product itself untouched.
What they asked
The instruction
Remove only the background. Do not alter the bottle, label, lighting, crop, colors, proportions, or reflections.
What the AI produced
The observable result
A clean background, but the label lettering changed, the cap warped, reflections moved, and the crop tightened.
Why
Why WWW assigns this verdict
The instruction was specific. The mismatch came from the editing method: a generative image model synthesized a plausible new image instead of isolating the original subject pixels.
What gave it away
Protected details changed together across the whole subject. That pattern points to regeneration, not one forgotten instruction.
Fix
What to do next
- Return to the original image; do not continue editing the regenerated version.
- Use segmentation or a background-removal tool when exact subject preservation matters.
- Compare the subject edges, lettering, proportions, and crop at full resolution before approving the export.
Model recommendation
Stay, retry, or switch?
Use a non-generative background remover such as Photoshop Remove Background or remove.bg for the cutout, then use generative editing only on the background layer if needed.
General lesson
What survives this one case
When exact pixels must survive, choose a tool that isolates pixels rather than one that redraws them from a description.