How to Remove Scratches from Old Photos with AI

The Problem with Scratched Photos

Every family has them — old photographs covered in scratches, creases, and scuff marks from decades of handling. Maybe the photo was carried in a wallet, stored loosely in a drawer, or shuffled between albums too many times. Whatever the cause, scratches are the most common form of photo damage and one of the hardest to fix by hand.

Types of Scratches and Surface Damage

Not all scratches are created equal. Understanding what you're dealing with helps set expectations for restoration.

Light Surface Scratches

These are fine, white lines across the photo surface caused by friction — sliding across a table, stacking prints without dividers, or rubbing against album plastic. They affect only the top coating and are the easiest for AI to repair because the underlying image data is still intact.

Deep Scratches

These cut through the emulsion layer and into the photo paper itself. You can often feel them with your fingertip. Deep scratches remove actual image data, so the AI must reconstruct what was underneath based on surrounding context. Results are usually excellent on backgrounds and clothing, and good on facial features.

Tears and Creases

Tears rip the photo apart. Creases fold and crack the emulsion without separating the paper. Both create harsh lines with displaced image data on either side. For tears, tape the pieces together on the back of the photo before scanning so the AI has a complete image to work with.

Pattern Scratches

Photos stored in textured albums sometimes develop a grid or diamond pattern embossed into the surface. These repetitive scratches are particularly challenging for manual editing but AI handles them well because the underlying pattern is predictable.

Why Traditional Methods Fall Short

Manual Retouching

Photoshop's clone stamp and healing brush tools can remove scratches, but it takes time and skill. A single heavily scratched photo might require 30-60 minutes of careful work. Multiply that by a stack of family photos and the project becomes overwhelming.

Automated Filters

Basic noise reduction and sharpening filters blur the entire image to hide scratches, but they also blur the details you want to keep. You end up with a soft, plastic-looking result that lost its character.

Professional Services

Sending photos to a professional retoucher produces excellent results, but costs $25-$100 per image with turnaround times of several days.

How AI Scratch Removal Works

AI restoration models have been trained on millions of image pairs: damaged photos alongside their clean originals. The AI learns to identify scratch patterns — their shape, direction, color, and texture — and distinguish them from intentional image content like wrinkles on a face or branches on a tree.

When you upload a scratched photo, the AI:

1. Detects the scratch locations — mapping every scratch, scuff, and crease in the image
2. Analyzes surrounding context — sampling color, texture, and detail from undamaged areas nearby
3. Reconstructs the hidden content — filling in each scratch with synthesized image data that matches the surrounding area
4. Blends the repair — ensuring the reconstructed areas have consistent lighting, grain, and texture

This entire process takes about 10-30 seconds.

Step-by-Step: Removing Scratches with PhotoFlip

1. Prepare Your Scan

Scan the photo at 300 DPI or higher. If using a phone camera, make sure there's no glare on the scratch lines — glare from overhead lighting makes scratches look white and prominent, which actually helps the AI detect them, but can also introduce false artifacts.

2. Upload to the Restore Tool

Go to the restore tool and upload your image. The AI automatically detects and processes all types of damage, including scratches.

3. Review the Result

Use the before/after slider to compare. Zoom in on areas where scratches were heaviest to check the reconstruction quality.

4. Run Face Restore if Needed

If scratches crossed over faces, the face restore tool can clean up any remaining artifacts in facial features. The face model is specifically trained on human features and does a better job on eyes, noses, and mouths than general restoration.

What to Expect from Heavily Scratched Photos

AI restoration is powerful but not magic. Here's a realistic expectation guide:

• Light scratches — removed completely, nearly invisible in the output
• Moderate scratches — removed with minor smoothing in affected areas
• Heavy scratches over backgrounds — excellent results, backgrounds are easier to reconstruct
• Heavy scratches over faces — good results, occasionally with slight softening of features
• Torn photos taped together — visible improvement but seam lines may remain faintly

Tips for Best Results

• Scan at the highest resolution your scanner supports — more pixel data gives the AI more to work with
• Don't pre-process scratches — avoid painting over scratches in Photoshop before uploading. The AI needs to see the original damage pattern
• Try the restore tool twice — for heavily damaged photos, running the restored result through a second pass can catch remaining artifacts
• Combine with upscaling — after restoration, upscale the result to recover detail that was lost in the damage

Start Removing Scratches Now

Upload your scratched photo to the restore tool and see the results in seconds. No signup, no software to install, and your photos are processed directly in your browser.