

Photo retake guide
One person blinked and everyone else looks great. Name who to fix and the retake changes only that face. Nobody else needs to pose again.


Every extra face is another chance that someone blinks, looks away or is mid-word when the photo is taken. In a big group, a frame where everyone looks good can be rare.
Instead of swapping a face in from a second photo, a retake works on the one photo you have. You name the person, and the prompt asks the model to change only that person's expression and to keep everyone else, the table and the framing as they are.
Hover or tap a photo to see Take 1. Each Take 2 is a real GPT Image 2.5 retake of Take 1. The people shown are AI-generated.








Name who to fix and the prompt asks for everyone else to stay as they were. Move the lens over the photo to see Take 1 underneath.

Move the lensThe prompt is written for you, and you can read it and change it in Create before anything runs.

Any photo you'd keep if it weren't for the face. It waits on your device until you continue.
Tap what's wrong. For a group photo, say who. That becomes your prompt.
Create opens with your photo attached, GPT Image 2.5 selected and the prompt filled in.
Every guide opens the same tool with the right fixes already picked.
Yes. Type who to fix, for example "the man on the left in the blue shirt". The prompt tells the model to change only that person and not to change anyone else.
Use position and clothing: left, right, middle, back row, plus a shirt colour or hair. The model cannot know people's names.
Fix one at a time. Retake the first person, then open that result in Create and retake the second person.
One retake of a standard 4:3 phone photo costs 21 credits. Other photo shapes cost 19 to 26 credits. This page shows the price for your photo before you continue, and Create shows it again before you press Generate.
Keep the moment. Fix the face.



