Proofsheet

Viral trend

Old photo restoration trend: 5 prompts to try

This trend uses AI to repair old family photos: scratches, creases, tears, stains and fading. The prompts ask the model to fix the damage only and leave every face exactly as it is.

Why it's trending

Many families have old prints that are torn, faded or stuck in albums. A cleaned-up digital copy is easier to share with relatives and keep safe, and the original stays untouched.

How to make it

  1. Scan the old photo, or photograph it flat in daylight without glare.
  2. Choose the prompt above that matches the damage, and set the finish in Customise.
  3. Copy the prompt, open ChatGPT, Gemini or Grok, upload the scan and paste it.
  4. Put the result next to the original and check every face, hand and background detail.

Tips

  • Scan at a high resolution and crop out the album page so the model sees only the photo.
  • Fix one problem at a time if the photo is badly damaged: tears first, then fading.
  • If a face looks smoother or younger than the original, ask the model to keep the original face exactly.
  • Keep the original scan as well; the restored copy is an interpretation, not a replacement.

Use photos with care

Only restore photos of your own family, or photos you have permission to edit. The AI fills in missing parts by guessing, so the result should not be treated as a true record in place of the original photo. Keep the original scan, and check every face against it.

Questions

Will the AI change my grandparents' faces?

It can. The prompts tell the model to keep every face exactly as it is, but always compare the result with the original and regenerate if a face looks different.

Can it rebuild a part of the photo that is missing?

It will guess what was there, and the guess can be wrong. Treat filled-in areas as an estimate, not as a record of what the photo showed.

Should I throw away the original after restoring?

No. Keep the original print and scan. The restored version is a copy made by AI and is not a replacement for the original.

Related trends

All trends

Prompt copied