How to Restore Old Photos: A Practical Guide

Start by making the problem smaller

Old photographs fail in several different ways at once, and it is much easier to fix them one at a time than to look for a single button that does everything.

Sort what you are looking at into four buckets:

• Fading and colour shift. Prints from the 1960s to the 1990s drift towards orange or magenta as the dyes break down at different rates.
• Physical damage. Creases, tears, missing corners, sticky album residue, fingerprints.
• Softness and grain. Either the original was taken on fast film, or the copy you have is a photograph of a photograph.
• Low resolution. Usually the scan or phone snapshot is simply too small.

The last two are what AI handles best. The first two need a little thought first.

Get a better copy before you get a better result

Whatever you do next is limited by the version you start from, and most people start from a bad one.

Scan, do not photograph, if you can. A flatbed scanner at 600 dots per inch gives you far more to work with than a phone held over a print. Most home printers include one, and libraries usually have one.

If you must use a phone, control the light. Put the print near a window, not under a ceiling light, and photograph it from directly above so the edges are not stretched. Turn the flash off. Flash bounces straight back off the print and blows out the middle.

Take the print out of the album or frame first. Glass adds reflections and album plastic adds a texture that the AI will faithfully treat as part of the photograph.

Photograph the whole print, edges included. You can crop later, and the border is useful evidence of the original colour if you want to correct the shift.

Restore it here

Upload the best copy you have. This runs the same enhancement used across Enhance AI, and it is free for three photos a day without an account.

The slider on the result compares it against what you uploaded, which is the fastest way to judge whether it actually helped.

Fixing the colour cast

If the whole photo has drifted orange, that is dye fade and it is very consistent across the image, which makes it one of the easier problems.

Any photo editor with a levels or curves control will fix it, including the free ones built into Windows Photos, Apple Photos and Google Photos. The trick most people miss: find something in the picture that you know was neutral grey or white, a shirt, a wall, a sheet of paper, and use the grey point picker on that. The whole image corrects around it in one move.

Do this before enhancing rather than after. Enhancement reconstructs detail based on what it sees, and it does a better job when the colours are already roughly right.

What to do about creases and tears

This is the part AI enhancement does not solve, and it is worth saying so plainly rather than letting you find out.

An upscaler treats a crease as a real feature of the image and will happily give you a sharper, higher resolution crease. Physical damage needs to be painted out, either by hand in an editor with a clone or healing tool, or with a dedicated repair pass.

The order that works: repair the damage first, correct the colour second, enhance and enlarge last. Enhancing first bakes the damage in.

Faces are the part people actually care about

In practice nobody examines the background of a family photograph. They look at the faces.

Face restoration is the strongest part of modern enhancement, because models have seen far more human faces than anything else, so their guesses about what a soft face should look like are unusually good. Eyes, hairlines and the edges of a smile tend to come back convincingly.

One caution worth keeping in mind. Restoration reconstructs, it does not remember. On a very small or very soft face it will produce a clear face that is subtly not quite the person you remember. If the result feels slightly wrong to you, trust that feeling, and keep the original alongside it.

Keep the original, always

Save the restored version as a new file and never overwrite the scan. Two reasons: tools improve, so a scan you restore again in three years will come out better, and the original is the only actual evidence of what the photograph looked like.

If these are family photographs, save the scans somewhere that is not one hard drive. The restoration is replaceable. The scan of a print that has since been lost is not.

Making it big enough to print

Restoring and enlarging are different jobs. Once the photo looks right, the AI image upscaler is what gets it to a size you can frame, and the image enhancer is the better tool if the problem is softness rather than size.

For a print, aim for roughly 300 pixels per inch at your intended size. An 8 by 10 inch print therefore wants about 2400 by 3000 pixels.

FAQ

What is the best way to scan old photos at home?

A flatbed scanner at 600 dots per inch is the best home option and is good enough for prints up to about 8 by 10 inches. Scan at that resolution even if the print looks small, because the extra pixels give any later enhancement much more to work with.

Can AI fix a torn or creased photograph?

Not on its own. Enhancement treats a crease as part of the picture and will sharpen it along with everything else. Repair the damage first in an editor using a clone or healing tool, then enhance, then enlarge.

Will restoring an old photo change how the person looks?

It can, and this is worth knowing before you restore a photograph that matters to you. The model reconstructs plausible detail rather than recovering the real detail, so on very soft or very small faces the result may be subtly different from the person you remember. Always keep the original scan.

Should I fix the colours before or after enhancing?

Before. Correcting a colour cast is straightforward and gives the enhancement accurate colour to work from, which produces a better result than enhancing first and correcting afterwards.

Is it free to restore old photos here?

Three a day are free with no account, no email and no watermark on the result. A free Enhance AI account adds credits if you have a larger batch to work through.

Related guides

• How to fix pixelated photos
• How to unblur an image