How to Remove Scratches from Old Photos (2026)

To remove scratches from an old photo, scan it at high resolution, run it through an AI restoration tool that repairs damage automatically, then check the result against the original. For most family photos that takes seconds and needs zero editing skill. This guide covers how to do it properly, when manual touch-up is still the better call, and the damage no tool can truly fix.

How AI scratch removal actually works

A scratch, crease, or dust speck is missing or wrong image data. Older manual methods cloned nearby pixels over the damage by hand. AI restoration does something different: it recognizes the damaged area and reconstructs what should be there based on the surrounding image and patterns it has learned — so it can rebuild across a face or a patterned background where simple cloning would smear.

That distinction matters for your expectations. The tool isn't uncovering hidden pixels; it's making an informed reconstruction. Most of the time that reconstruction is excellent. Occasionally it guesses wrong, which is why you always compare to the original.

Step by step

1. Scan high, not low. Repair quality is capped by scan quality. Scan at 600 DPI or higher and save as TIFF or PNG, not JPEG. If you're scanning a batch, our guide to scanning old photos covers settings and handling.
2. Clean the surface first. Gently blow or brush loose dust off the print and wipe the scanner glass. Real dust you remove now is detail the tool doesn't have to guess at later.
3. Remove scratches before sharpening. Run the damage-repair pass first so sharpening works on a clean image instead of crisping up the scratches.
4. Compare at 100%. Zoom in on faces and any detailed areas — fabric, hair, lettering. This is where a tool's reconstruction either holds up or reveals itself.
5. Re-run or touch up the misses. If one area smeared, lower the strength and run again, or fix that spot by hand.

When manual touch-up still wins

AI is the right default for almost every old family photo. Reach for manual editing — a clone or healing brush in an editor — in three cases: a severely torn one-of-a-kind print where you want a guaranteed, hand-checked result; a spot where the AI keeps reconstructing the wrong detail and you need to override it; or when you want total pixel-level control over a specific area. For a single irreplaceable print, it's also reasonable to pay a professional restorer. To weigh the automatic tools against that, see our comparison of photo restoration tools.

What no tool can fix

Be honest with yourself about limits. If a scratch destroyed an eye entirely, the tool will invent a plausible eye — not recover the real one. Detail that was never captured (a face lost to a deep crease, a corner torn away) gets reconstructed, not restored. That's usually fine for a keepsake, but if accuracy is critical — a legal document, a genealogical record — note where the image was rebuilt.

If your photo is blurry rather than scratched, that's a different fix; see fixing blurry photos. And once scratches are gone, the full restore-old-photos workflow covers upscaling and colorizing.

FAQ

Can AI remove deep scratches and tears? Most surface scratches, creases, and dust, yes. Deep tears through a face are reconstructed rather than recovered, so check against the original.

Is AI scratch removal better than Photoshop? For most family photos, yes — seconds, no skill needed. Manual editing wins when you need pixel-level control over an irreplaceable print.

How do I remove scratches without losing detail? Start from a high-res scan, repair before sharpening, and compare at 100% — lower the strength if real detail smeared.