№ 74 · RESTORATION

Iterative ChatGPT restoration: repair, de-smooth, then verify

№ 74

№ The prompt

Restore this scan across three passes, and run them one at a time. Pass 1: Repair scratches, creases, dust, and fading. Recover detail that is genuinely present and keep every face unchanged. Pass 2: The last result looks over-smoothed — reduce the smoothing and add back a subtle, natural film grain that matches the era. Pass 3: Compare this result against my original scan and undo any change to the faces or features; the people must still clearly look like themselves. Tell me anything you could not recover.

Open ChatGPT

№ How to use it

  • Input

    A scan uploaded into ChatGPT, kept available across the whole edit session for the comparison pass.

  • Preserve

    Faces, grain, and composition across every pass; the original stays the reference.

  • Avoid

    Accepting over-smoothing, letting faces drift between passes, or regenerating a new image.

№ 03

Repair scratches and creases, recover faded detail, add period-accurate color while keeping the original grain.

Restore and colorize a damaged old family photo

RESTORATION

Repair scratches and creases, recover faded detail, add period-accurate color while keeping the original grain.

№ 07

Recover edge detail from a blurred capture, sharpen the face that is already there, suppress noise without plastic skin.

Sharpen and deblur a motion-blurred snapshot

RESTORATIONAI

Recover edge detail from a blurred capture, sharpen the face that is already there, suppress noise without plastic skin.

№ 12

Neutralize the aged yellow cast, rebalance contrast, recover shadow detail, keep the photo's original character.

Remove yellow tint and fading from a scanned print

RESTORATION

Neutralize the aged yellow cast, rebalance contrast, recover shadow detail, keep the photo's original character.