Bringing old family photos back to life
Scratches, fading and yellowing can hide the faces we care about most. Here's how we approach restoring them.

Most families have a box of old prints somewhere: creased, faded, yellowed, sometimes torn. They are often the only pictures we have of the people in them.
We built Restore AI to make repairing those photos something anyone can do in a minute, from their phone.

One problem at a time
Old photos rarely have just one problem, so Restore AI is built from separate tools that each do one job:
- Descratch removes scratches, dust, folds and small tears.
- Face Restoration brings back detail in faces, which is usually what matters most.
- Colorize adds natural colour to black-and-white photos.
- Deblur and Clarity Upscale sharpen soft or low-resolution prints.
- Yellow Tint Removal corrects the colour cast that old prints develop over time.
You can use one tool, or combine them.
Colour is a reconstruction
When an old photo is colourised, the model is making an informed guess about what the colours were — skin, sky, foliage and fabric. The results can be remarkably natural, but they are an interpretation. If you know that grandad's suit was navy, trust your memory over the model.

Tips for the best result
- Scan or photograph the print flat, in even daylight, without glare.
- Fill the frame with the photo so the app has as much detail as possible.
- Keep the original. Restoration is a copy, never a replacement.
Restore AI is available on the App Store and Google Play in 38 languages. See Restore AI.
More from the blog
All posts
AI connected like neurons
Why we build small, focused AI apps that share the same models underneath.
Read post
How AI upscaling works — and when it doesn't
What actually happens when an app turns a tiny photo into a large, sharp one, and how to get the best results from it.
Read post
Blur before you share
A practical guide to protecting faces, licence plates and documents in the photos and videos you post online.
Read postQuestions or ideas about this post?
Write to us — every message is read by a person on our team.