What the tech

Search

/ opens search from anywhere. Esc closes it.

Tools

Get the newsletter
Example result of this prompt The made-up input photo InputExample result

Example Example made on our own Mac with open models (Z-Image and FLUX.2 klein). Your result in the AI above will look different. How we make examples

Restore an old photo with scratches and creases

Remove scratches, creases, stains and fading from an old print, without changing who's in it. You need a scan or photo.

You need
1 photo of yours
Written for
Gemini, ChatGPT
Result
Image

Prompt

Restore my attached old photograph. Remove the scratches, creases, tears, dust spots, stains and fading, and fill in any missing corners so they match the surrounding image. Bring back gentle contrast and clean up the tones, but keep it color and keep the natural film grain and the look of the era. Do not add details that weren't there, and do not modernize the clothes, hair or setting. Keep every person's face exactly as it is in the photo: same identity, features, age, expression and proportions, with no smoothing or beautifying.

Fill in, or keep the brackets and change them later

color

How to use it

  1. Scan the print, or photograph it flat in daylight with no glare.
  2. Open Gemini or ChatGPT, choose the image model, attach the photo and paste the prompt.
  3. Something changed? Ask it to fix only the damaged area and leave the rest exactly as it was.
What to expectHeavily damaged areas are reconstructed by guesswork, so faces can still change a little, and small details may be invented.

Tips

  • Keep the original scan: the restored copy is an interpretation, not a record.
  • Badly damaged faces are guessed; compare them with other photos of the same person.
  • Get the family's OK before uploading photos of relatives.

Editor: Ali Dezhami · Updated November 18, 2026