How to Restore Old Photos: Repair Faded and Damaged Pictures

Last updated: July 18, 2026

Old photos fade, crack, and gather dust. Restoring one means getting a clean high-resolution scan, repairing the physical damage digitally, and using an AI enhancer to rebuild the detail that time erased. This guide walks through the whole sequence — from how to scan a fragile print without damaging it to the exact order for fixing tears, fading, and grain.

Quick answer: how do you restore an old photo?

Scan the original at 600 DPI or higher, save it as a lossless TIFF or PNG, then use an image upscaler to rebuild detail and an AI photo enhancer to clean up fading and grain. Fix large tears and stains manually before running the AI pass, because the model follows what it sees — if a crease runs across a face, the AI can hallucinate features. Always keep the untouched scan separate from the edited version.

Why do old photos degrade?

Photographic prints break down through three main forces: chemical fading of the dyes, physical damage from handling and storage, and light exposure. Color prints from the 1970s and 1980s are the most fragile because their dyes were unstable. Black-and-white silver gelatin prints hold up better, but the paper itself yellows and becomes brittle.

Knowing the failure mode tells you the fix. Chemical fading needs color correction and contrast rebuild. Tears and creases need cloning or AI inpainting. Mold and staining need selective cleanup. The Northeast Document Conservation Center publishes guidance on preserving photographic materials that is worth reading before you handle originals.

How should you scan an old photo?

The scan is the foundation. A poor scan cannot be saved later, so get this right.

Source type	Minimum DPI	Format	Notes
Standard print (4×6)	600	TIFF / PNG	Captures all printable detail
Small wallet photo	1200	TIFF / PNG	Lets you enlarge later
Slide or negative	2400-3200	TIFF	Needs a transparency-capable scanner
Fragile or torn print	600	TIFF	Scan flat, no pressure on lid

• Clean the scanner glass with a microfiber cloth first; dust scans in as permanent specks.
• Place the photo flat without forcing the lid down — bend damage scans poorly.
• Scan in color even for black-and-white photos; you keep the option to convert later.
• Save the master scan as TIFF or PNG, then make a working copy for editing.

Step-by-step: restore a faded photo

Work in this order. Each stage depends on the one before it.

1. Scan at 600 DPI and save a lossless master you never edit directly.
2. Crop away borders and scanner edges with the image cropper.
3. Repair large tears and creases with a clone or heal tool — do this by hand for faces.
4. Correct the color cast: sample a region that should be neutral gray and shift white balance.
5. Rebuild contrast by lifting the black point and pulling highlights down.
6. Run the image through an AI upscaler to rebuild lost facial detail, or the AI Face Restoration tool if the face itself is blurry or low-detail rather than just soft.
7. Reduce noise last, after the AI pass, so you do not erase real detail.

If the photo is badly damaged, do steps 3-5 on a duplicate layer so you can compare against the original scan. The old photo restoration guide goes deeper on the AI rebuild stage, and the AI face restoration guide compares GFPGAN against CodeFormer for step 6 specifically.

When is manual repair better than AI?

AI restoration is fast but not careful. It guesses at missing detail, and on faces that guess can be wrong — a relative's features subtly changed, an eye rebuilt differently. For heirloom photos where accuracy matters, do the structural repair by hand and use AI only for texture and resolution.

• Hand-repair tears, holes, and missing chunks — the AI will fill them with plausible but invented content.
• Use AI for overall sharpening, denoise, and resolution upscaling where guessing is safe.
• Always compare the AI output side-by-side with the manual version on faces.
• If a face looks "too good" after AI, it probably is — dial the strength back.

What can AI restoration not fix?

Some damage is beyond clean repair, and recognizing that saves wasted effort.

Damage type	Can AI help?	Reality
Fading and color shift	Yes	Color rebuild works well
Overall grain and noise	Yes	Denoise is reliable
Tears and creases	Partially	Hand repair is safer
Completely missing sections	No	AI invents content
Water or mold damage on faces	Risky	Likely hallucinated
Blurry original	Limited	Upscaling adds pixels, not focus

Exporting and sharing the restored photo

Once the restoration looks right, export for the destination.

1. Keep a full-resolution PNG master as your archive.
2. For sharing online, resize to the target platform and compress to keep the file light.
3. For re-printing, leave it at the scan resolution and export as PNG to avoid JPEG banding.
4. For a family archive, make a copy at reduced resolution for email.

• JPEG re-saves compound artifacts, so convert to your final format once.
• WebP holds grain detail better than JPEG at the same size — see the format comparison.
• If you plan to print large, the resolution guide explains DPI targets.

Choosing between manual and AI tools

The right tool depends on how many photos you have and how damaged they are. A single heirloom portrait deserves a careful manual pass. A box of three hundred faded vacation prints calls for a faster workflow.

• One important photo: scan, hand-repair the damage, use AI only for sharpening and upscale.
• A dozen family photos: AI-first pass on all of them, then manual touch-ups on the few faces that look wrong.
• A large archive: run a batch process with the same AI preset, accept the trade-off in accuracy for volume.
• Slides and negatives: use a dedicated film scanner at high DPI, then denoise — the noise reducer guide covers the settings that work on film grain.

The mistake to avoid is treating AI restoration as a one-click fix on irreplaceable photos. It is fast, but it is not conservative. For anything you cannot re-scan, keep the master scan and every intermediate step so you can roll back.

Caring for the originals

Restoration does not protect the physical print. Store originals in acid-free sleeves, keep them out of direct light, and avoid damp basements or hot attics. The Library of Congress offers practical advice on preserving family photographs that covers storage materials and environment. A restored digital file is only as good as the original it came from, so scan the important prints now while they are still legible.

Restoring an old photo is mostly patience and the right order of operations: scan high, fix structure by hand, let AI handle texture, and export for the destination. The reward is a picture that lasts longer than the paper it was printed on.

Frequently asked questions

Can AI restore any old photo?

For denoise, upscale, and scratch repair, yes. For faces and colorization, AI guesses — plausibly but not accurately. Treat AI as a strong first pass on the safe jobs and keep the original scan as the source of truth.

What DPI should I scan an old photo at?

At least 600 DPI as TIFF. The scan resolution caps everything downstream; AI cannot recover detail the scan never captured. A low-DPI scan limits the restoration no matter how good the model.

Should I colorize an old photo?

Only if historical accuracy is not the goal. Colorization is an interpretation, not a recovery of the original colors. For a family album, fine — the Old Photo Colorizer adds plausible color in one pass — but for anything documentary, keep the unmodified scan and label any colorized version as an interpretation.

When is manual repair better than AI?

For tears, holes, and missing chunks, where you reconstruct structure rather than guess texture. AI fills small gaps plausibly but fails on large structural damage. Repair tears by hand first, then let AI handle denoise and texture.

Can I restore a photo on my phone?

Basic fixes (crop, exposure) yes, but serious restoration (scratch repair, AI denoise, upscale) needs a desktop editor and a high-DPI scan. Phone apps are limited; the scan quality matters most. See the restoration guide.

Can AI restore a torn photo?

Partially. AI fills small gaps and tears plausibly, but large structural damage (a missing corner, a torn face) needs manual reconstruction first, then AI for texture. Repair tears and holes by hand before AI touches the file, because AI on a gaping hole invents structure rather than recovering it. Manual repair first, AI denoise and texture second.