CraterView.ai model

Restore: AI photo restoration for old, damaged prints

Repair an old photo. Restore removes tears, creases, scratches, dust and grain and corrects faded colour, and returns a black-and-white photo in colour or keeps it black and white.

Open Restore in the web app Use it from your code

No sign-up and no card — a workspace opens in your browser the moment you arrive.

A print in a shoebox has been through decades of handling: a fold, a torn corner, white specks where the emulsion flaked, dyes drifted to amber. A scanner records all of that faithfully. Restore takes away what the years added and gives you back the picture of the day. Scan the print, drop the scan in, and nothing needs retouching by hand.

Two young men in dark jackets seated side by side, one with a beard, in a scratched and stained sepia tintype — A scratched and stained tintype, repaired and returned in colour. Drag the divider to compare.

The same photograph after Restore, the scratches and stains gone and the two men in natural colour, in brown jackets and striped trousers — A scratched and stained tintype, repaired and returned in colour. Drag the divider to compare.

What it does

Restore repairs the whole photo at once, and returns it at its own size: it is a restorer, not an enlarger. The colour it puts back is usually right about wood, sky and wool:

Nikola Tesla reading in his Colorado Springs laboratory beneath arcs of electricity from a giant coil, in a black-and-white photograph from 1899 — Nikola Tesla’s Colorado Springs laboratory, 1899. Photograph by Dickenson V. Alley.

The same photograph after Restore, in colour, the wooden walls and rafters warm brown, the coils dark, and the discharge still arcing across the room — Nikola Tesla’s Colorado Springs laboratory, 1899. Photograph by Dickenson V. Alley.

A scan of 300 to 600 dpi is plenty; a higher setting records more of the paper’s texture without giving the repair anything more to work with. Send the TIFF your scanner software saves, or a PNG or a JPEG. A 16-bit scan keeps all its tones through the repair, and a TIFF comes back as a PNG, which holds them.

Ways to run it

In colour, or black and white

With monochrome off, a black-and-white photo comes back in colour. Turn it on to keep it black and white, with a gentler repair. When the colour is not yours to guess — a uniform, a flag, a car — don’t ask for it.

Albert Einstein at a blackboard in a faded sepia photograph — Albert Einstein lecturing in Vienna, 1921, with **monochrome** on: the repair, without the colour. Photograph by Ferdinand Schmutzer.

The same photograph repaired and kept black and white, the fading and the specks gone — Albert Einstein lecturing in Vienna, 1921, with **monochrome** on: the repair, without the colour. Photograph by Ferdinand Schmutzer.

Tell it what the photo can’t show

A black-and-white photo holds no colour to restore, so Restore draws a plausible one. Write one sentence in context about what you know of this photo: who is in it, their age, the colours they wore, where and when it was taken. The repair follows it. Describe the photo rather than naming someone famous, because a famous name pulls the result towards that person’s best-known pictures. Below, left to right: the original, the default, a context naming him, and a context describing him.

The same face four ways: the faded 1921 original; the default restore with grey hair; with a context naming Albert Einstein, an elderly man with white, wild hair; with a context describing a man in his early forties with dark, greying hair, closer to the original

Skin tone is where this matters most. Colourizing a black-and-white photo can lighten darker skin, and a sentence stating the subject’s skin tone corrects it. The repair follows the whole sentence, so write only what you know, and compare the background with the scan.

A different repair

The seed setting picks one repair out of the many Restore could make. The same photo and seed always give the same result, and a different seed gives a different repair, which is worth trying on a photo that matters.

Very small originals: Enhance first

A scan a few hundred pixels across gives Restore little to repair from. Enlarge it with Enhance first, then restore the result. It is two runs, so only chain them when the original is genuinely small.

A steamship with tall masts and a smoking funnel, and a small tug beside it, on calm harbour water in a faded black-and-white photograph — A photograph 301 pixels wide, enlarged with Enhance and then restored.

The same photograph enlarged with Enhance and then restored, in colour, the white steamship with its dark hull, the tug's red-brown cabin, and blue water and sky — A photograph 301 pixels wide, enlarged with Enhance and then restored.

What to check

Faces and hands. The repair repaints the whole photo, and a face that is small in the frame can shift a little. Compare the two at the size you will use the photo.
Writing and fine pattern. A caption on the print, a shop sign, a check shirt: these are where a repair is most likely to show.
Colours that carry meaning. Uniforms, insignia, flags and vehicles are drawn, not recovered.
Keep the original. Say that a picture was restored, and that its colour is a rendering.

From your own code

Restore is on the HTTP API, with clients for Python and TypeScript. Below every run in the web app, Equivalent HTTP API calls shows the same request in curl, Python and TypeScript, ready to copy. Your first API call, in three requests walks through getting a key and the three requests a job takes.

Try it on your own file