How to restore old photos with AI: we tested it against the originals

Arthur Down — Editor, N33 blog
Published November 5, 2026

Upload a scan, pick Restore or Restore and colorize, get a repaired photo. We aged four photos ourselves and checked every result against the real original.

To restore an old photo with AI, scan the print or photograph it flat in even light, add it to our photo restoration tool, pick Restore or Restore and colorize, and you get a repaired version in a minute or two. We aged four photos ourselves, so every result below can be checked against the real original.

A note before the comparison: we make one of these tools, and it is the only one we tested ourselves. For the others, every claim below comes from their own pages, linked where it appears and checked the day this guide went live.

What we measured ourselves
2026-11-05 → 2026-11-05
10 of 12restored pictures that kept the person's face, four aged photos, Restore and Restore and colorize

We took four colour portraits of adults (a woman with long straight hair, a woman with coily hair, a man with wavy hair and stubble, a man in a denim jacket), turned two sepia and two black and white, and added scratches, dust, light leaks and a folded corner. Each photo got both looks, and the two photos of the first woman and the man with wavy hair got both looks a second time: 12 runs, one picture each, 21 to 98 seconds. Every scratch, spot and light leak was gone in 12 of 12. Ten kept the face; in the first run of the woman with long straight hair both pictures came out with a different-looking face and her colorized picture got a new background, while her second run kept her face. Restore and colorize gave full colour in 6 of 6, after we rewrote that prompt to ask for colour everywhere with no sepia left. Restore kept the sepia of a sepia print in 1 of 3 runs and turned the other two neutral grey.

Source: N33 photo restoration on Flux Kontext Max (Pro), staff run; every picture compared by eye with the photo before we aged it

Can AI really restore old photos?

Yes, for the usual damage of age. In our test it removed every scratch, dust spot and light leak, and kept the face in ten pictures of twelve. It redraws what it repairs, so a face or a background can come back changed: compare the result with the print.
Top: three of the photos we aged. Bottom: Restore. The scratches, dust and light leaks are gone.

How do you restore old photos with AI?

Scan the print, or photograph it flat by a window without glare. Upload it, choose Restore to repair it in black and white or Restore and colorize to add colour, then compare the result with the print and run it again if a face looks different.

Can AI colorize a black and white photo?

Yes. Restore and colorize repairs the damage and adds natural colour to skin, hair, clothes and the background. The colours are a likely guess, not a record: compared with our originals, skin and hair were close, while clothes and backgrounds often were not.
Top: Restore and colorize. Bottom: the real photos before we aged them. Skin and hair come close; a mustard sweater turned grey.

How accurate is AI photo restoration?

The damage goes every time; the face does not always stay. Ten of our twelve pictures kept the person. The two that did not were one woman's first run, and her second run kept her face, so if someone looks different, run the photo again.

Which prompt works for AI photo restoration?

Tell the model to keep the same photograph: the faces, poses, framing and people. Then list the damage to repair. Each picture below has a Copy prompt button with the exact prompt our tool sends, run on Flux Kontext Max, an image editing model.
A restored black and white portrait of a smiling woman with coily hair under trees, no scratches left
Restore: the damage gone, sharp detail back.
A colorized portrait of a man in a denim jacket and white T-shirt on a brick street
Restore and colorize: full colour, the face kept.

Which AI photo restoration tools can you try?

MyHeritage repairs a few photos free and asks for a Complete subscription after that. PicWish gives a few free uses and then credits. Ours is part of Pro and runs Flux Kontext Max, with or without colour.
AI photo restoration tools compared, from each service's own page, checked 5 November 2026
ServiceFreeWhat it does
N33No, part of ProRestore, or Restore and colorize, on Flux Kontext Max; the photo is deleted when the run ends
MyHeritageSeveral photos free; more need a Complete subscriptionPhoto Repair, inside a family history site
PicWishA few free uses, then creditsRemoves scratches and creases, and colorizes

Our row is from our own test; the others are as each service describes itself, linked here.

How do you get the best result from an old photo?

Start from the sharpest copy you have. Scan it, or photograph it flat with the phone straight above, in daylight and without flash or glare. Crop off the album page, and run it more than once: each run differs, and one is usually best.

To try it, open our photo restoration tool. For a family photo turned into a different decade instead, see our guide to AI yearbook photos, and to see someone older, read how realistic an AI age filter is.

Questions and answers

Is there a free way to restore old photos with AI?

MyHeritage repairs several photos free and PicWish gives a few free uses. Ours is part of Pro.

Will AI change the faces in my old photo?

It can. Our prompt tells the model to keep every face exactly as it is, yet in our test two pictures of twelve came back with a different-looking face. Compare with the print and run it again.

Are the colours in a colorized photo real?

No, they are a likely guess. Skin and hair usually come out close; clothes and backgrounds can differ from the real ones.

Can I restore a photo of a child?

Not on ours. Every face goes through an age check, and a photo that shows a child or a teenager is refused, family photos included.

What happens to the photo I upload?

On ours, it is used only for the run and deleted when the run ends; the restored pictures stay in your Generations until you delete them.

Arthur Down

Arthur Down is a journalist with more than 12 years of experience writing about technology, personal finance, health and everyday consumer topics. He studied Journalism and Media Studies at the University of Sheffield, started as a reporter at a regional newspaper and has worked freelance since 2014. At N33 he writes and edits the blog's practical guides to AI tools.

How we research, source and review articles