How to Restore Old Photos: Fix Scratches, Fading & Damage with AI

An old family photo can have several problems at once: scratches across a coat, faded faces, dust, grain, and too little detail for a larger print. Learning how to restore old photos starts with telling those problems apart. A cleaner image is only useful if your grandmother still looks like your grandmother.

You can restore old photos online without Photoshop experience. AI tools can reduce some damage automatically, but each tool has limits. This guide explains what to try, when to stop, and how to check that a restoration preserves the people and details you care about.

The cover and damage examples are AI-generated editorial illustrations, not PixResolve processing results.

Quick Answer: How to Restore Old Photos

1. Scan or photograph the original carefully. For prints, start around 300 DPI; use a higher optical resolution when you need an enlargement.
2. Identify the damage types using the table below. One photo may need several different repairs.
3. Choose a matching restoration tool. Check its supported repairs, file limits, and price before starting.
4. Compare the completed result at 100% zoom, especially faces, writing, and fine textures.
5. Download the full-size result and keep the original scan untouched.

Processing time varies with the image, tool, and connection. Allow time for scanning and checking, too; the automatic repair is only part of the job.

First, Diagnose: What Type of Damage Is It?

To restore damaged photos, inspect both the whole image and a close-up. Thin white lines suggest surface damage. Uniform softness may come from the original exposure, a poor scan, or low resolution. The repair should match the cause.

Illustrative damage patterns. Use the table to separate repairable distractions from missing information.

That last distinction matters most. A scratch crossing a recognizable cheek gives a restorer surrounding information to work with. A face scraped away entirely does not. Another family photograph may help guide a reconstruction, but the result should still be labeled as reconstructed.

Traditional Restoration vs. AI Restoration

A human retoucher can make local decisions: remove a scratch while leaving a nearby wrinkle, correct a color cast, or consult another portrait before rebuilding a damaged feature. Ask for a quote based on the actual image and scope; there is no useful universal price for every damaged print.

AI photo restoration estimates cleaner content from surviving information and learned patterns. The research paper Bringing Old Photos Back to Life treats scratches and dust differently from noise and blur. That helps explain why mixed damage needs different kinds of repair, though individual tools vary.

Automation can reduce manual work, but its output remains an estimate. Use it for a first pass when useful structure survives, then judge whether the important details stayed faithful.

How to Restore Old Photos, Step by Step

Step 1: Make a clear scan before editing

Start with the original print if you have it. Clean the scanner glass, include the full photograph, and avoid automatic enhancement when making your archive copy. For an ordinary print, 300 DPI is a practical starting point; consider 600 DPI if you plan to double its printed dimensions. Epson's scan-resolution guidance explains this relationship. Tiny prints and negatives may need different settings.

Scanner menus often say DPI; the useful measure in the saved file is pixels. A 4 × 6-inch print scanned at 300 pixels per inch produces about 1200 × 1800 pixels. Changing the file's DPI label afterward does not capture more detail.

Without a scanner, photograph a stable print in bright, even light, with the camera directly above it. Avoid flash, glare, and shadows. Do not force a brittle print flat or feed it through an automatic scanner. The National Archives' digitizing guidance explains how scanning can damage fragile originals.

Step 2: Repair visible damage, then inspect what remains

Choose scratch removal for surface lines, denoising for distracting speckles, or tone correction for fading. Start with the problem that obscures the subject most. If a tool combines repairs, review that first result before adding more processing; separate passes are useful for control, not a guarantee of better quality.

Color correction and colorization are different. Correcting a yellow cast adjusts existing color; colorizing a black-and-white print introduces estimated colors. Adobe's Neural Filters documentation lists Colorize separately from Photo Restoration and notes that fading can reduce color-prediction accuracy.

Leave enlargement until you have a restoration worth keeping. If the final print needs more pixels, use an AI image upscaler and inspect the enlarged version again.

Step 3: Check the face at 100% zoom

Compare the source and result at the same zoom. At 100%, one image pixel maps to one viewer pixel; also check the intended display or print size.

• Eyes and mouth: do their positions, shapes, and expression still match?
• Face shape and hairline: has cleanup changed a familiar identifying feature?
• Skin and clothing: did wrinkles, grain, or fabric turn into smooth patches?
• Writing and small objects: are dates, jewelry, and uniform details still supported by the original?

A face can look sharper and less like the person. If that happens, return to the original and try a gentler setting where available, a different method, or a retoucher. Repeatedly processing the altered version can make comparison harder.

Step 4: Keep the original untouched

Save the unedited scan, restored working copy, and sharing or print export separately. Keep the physical photograph, too. For backups, the National Archives recommends three copies across two types of storage, with one copy elsewhere.

Record names, approximate dates, and any reconstructed areas alongside the files. Future family members should be able to distinguish the surviving photograph from your edits.

Step 5: Export for the size you actually need

Download the full-size result rather than a screenshot of the comparison view. Keep a lossless working copy where possible, or use a high-quality JPEG export and avoid repeated recompression.

Check the pixel dimensions against your intended print size. When sharing, send the original exported file if the service offers that option; messaging apps may compress photos. A larger file can still contain altered details, so review it before ordering a print.

When AI Restoration Won't Save a Photo

Some damage leaves too little evidence for a faithful repair:

• Large missing sections: a filled corner may look plausible without matching the original scene.
• Destroyed faces: neither AI nor a human can guarantee an accurate likeness from information that is gone. Other portraits can help guide a clearly disclosed reconstruction.
• Water, heat, or surface damage: stains and lost emulsion may obscure the image beyond reliable recovery. Digital editing also cannot stabilize a deteriorating physical print.

For a fragile or irreplaceable original, ask a photograph conservator about handling and preservation. A digital retoucher can work on its scan. These are different jobs; cleaning up a file does not preserve the object itself.

Common Mistakes When Restoring Old Photos

• Accepting an invented face. Check identifying features, not just apparent sharpness.
• Removing all texture. Film grain and wrinkles can belong to the photograph; waxy smoothing is not automatically an improvement.
• Starting from a poor copy. Rescan if glare, compression, or missed focus hides details visible in the print.
• Repeating processing without comparing. Save each candidate separately and keep the original as your reference.
• Treating vivid color as historical evidence. An AI-colored dress or uniform may be convincing and wrong.

How PixResolve Fits This Workflow

PixResolve's AI photo restoration tool focuses on dust, fine scratches, and small surface marks. Upload one JPEG, PNG, or WebP, review the displayed access requirement, and start restoration. Compare the completed result before downloading it. This tool does not colorize black-and-white photos or promise to rebuild large tears.

The one-time Guest Trial can cover one eligible result without signup, with a source up to 2 megapixels. A high-resolution archive scan may exceed that limit; keep it and use a separate eligible copy if trying the Guest Trial.

Credit-funded restoration costs 15 Credits per result. Newly verified accounts receive 35 welcome Credits, valid for 30 days: enough for two restorations with 5 Credits remaining. Credits are reserved when you start processing; the comparison view appears afterward. Check the displayed cost before each new attempt.

Photos are stored privately for processing. Source and result access expires after 24 hours, with deletion handled under the Privacy Policy. PixResolve does not use customer images to train its own models; the policy also explains the external processing providers. Download files you want to keep promptly.

Frequently Asked Questions

Can an old photograph be restored?

Often, at least partly. Fine scratches, dust, fading, and noise may improve with suitable tools. Large missing sections and destroyed faces need reconstruction, whose accuracy cannot be established from the damaged photo alone.

Can I restore old photos for free?

You can try one eligible result with PixResolve's Guest Trial without signup, using a source up to 2 megapixels. Newly verified accounts get 35 welcome Credits for 30 days, enough for two 15-Credit restorations. The Restore page processes one photo at a time; this is not an unlimited free service.

Will AI restoration change how people look?

It can. Automatic cleanup may alter eyes, mouths, skin texture, or expressions even when the result looks believable. Compare the restored face with the original at 100% and at its intended display size. Reject changes that make the person less recognizable.

Should I restore old photos myself or hire a professional?

Try AI on a copy when the damage is light and the people remain recognizable. Consider a skilled retoucher for extensive damage or a portrait that needs careful local work. Ask a conservator about a fragile physical original, and request a quote for the specific job.

Should I scan or photograph the old photo?

Scan a stable, flat print if you can. Start around 300 DPI for an ordinary print, increasing the optical resolution for enlargement. A carefully lit phone photograph is an alternative when scanning is unavailable or the item cannot safely lie flat.

Keep the Photo Recognizable

Knowing how to restore old photos means recognizing which marks can be reduced and which details are missing. Start with a clear scan, choose a repair that fits the damage, and check faces before keeping the result. A modest improvement that preserves someone's expression is more useful than a dramatic reconstruction you cannot trust.

For blur caused by camera or subject movement, read our guide on how to unblur a photo. For broader noise, light, and color adjustments, explore the AI image enhancer.