How to Restore an Old Photo With AI

To restore an old photo with AI, scan the print at a high resolution, save that scan untouched, then work on a copy: clean up dust and scratches first, correct fading and contrast, apply AI enhancement, and finally compare faces carefully with the original. AI is very good at fixing the common signs of age, but when part of the image is missing it invents a replacement, so the best results come from AI plus your own careful review.

This guide is about old prints and scanned film. If you are working with a modern digital photo that is blurry, grainy or small, the related guides are linked further down.

What AI Restoration Can and Cannot Fix

The rule underneath all of this: the more information has been lost, the more the AI is guessing. A faded but intact print can be restored very faithfully. A face with half its features scratched away will come back as a face, not necessarily their face.

Step 1: Make a Good Scan

The quality of your scan sets the limit for everything that follows. A sharp, high-resolution scan gives the AI real detail to work with; a blurry phone snapshot under a lamp gives it very little.

Use a flatbed scanner if you can

• Clean the glass with a microfibre cloth. Every speck on the glass becomes a spot on your scan.
• Handle the print by its edges and brush off loose dust gently. Don't wipe a print with water or cleaning products.
• Scan at a high resolution. A common recommendation is 600 dpi for standard-size prints and higher for small prints you plan to enlarge. At 600 dpi, a 6 × 4 inch print becomes roughly 3600 × 2400 pixels, plenty for most uses. Dots per inch explains how scan resolution relates to print size.
• Turn off the scanner's automatic "enhancement" if you can, so you capture a neutral original and make the decisions yourself.
• Save the master scan as PNG or TIFF, not a compressed JPG.

If you only have a phone

Place the print flat in soft, even daylight, avoid overhead lights that cause glare, and hold the phone exactly parallel to the print. Fill the frame and avoid zoom. Phone scanning apps that correct perspective help.

Step 2: Protect the Original

Before any editing, set up a simple version structure:

Original print → Master scan (never edited) → Working copy → Restored version(s)

Never overwrite the master scan. If an AI result disappoints you next year, or a better tool appears, you can start again from it. For important photos, keep a copy of the master scan in a second place, such as a cloud folder or external drive.

Step 3: Crop and Straighten

Use Crop to remove the scanner bed around the print and Rotate if it is sideways. Keep damaged borders for now if you want the restoration to retain the print's original shape; trimming them is a style choice.

Step 4: Clean Up Dust, Spots and Scratches

Fix physical damage before you sharpen or enlarge anything. Sharpening makes dust and scratches more visible, and enlarging makes them bigger.

• Small dust spots and specks: the Object Remover lets you brush over a mark on your own photo and fills the area from its surroundings. Work on one spot at a time and zoom in to check the fill blends in.
• Scratches across backgrounds are usually easy to fill.
• Scratches across faces, hands or text need more care. Remove them in short sections and compare with the original after each one, so the fill doesn't change a feature. For printed date stamps or captions across the image, see how to remove text from your own photos with AI.

Step 5: Run AI Restoration

With the damage cleaned up, run the overall restoration. The Image EditPro Old Photo Restoration tool uses a neural network to sharpen, reduce noise, balance colour and enlarge the image to 2× in a single pass. It runs entirely in your browser, so family photos don't leave your device.

After it finishes:

1. Compare it side by side with your working copy at normal size.
2. Zoom in on faces, edges and repaired areas.
3. Check the background for strange patterns, warped objects or invented textures.

If the result looks over-processed, you can stop at the cleaned-up version from Step 4 and apply only gentle tone corrections instead.

If you need it bigger still

For a large print, you may need more than the 2× enlargement. Upscale the restored image with the AI Image Upscaler, and check faces again afterwards. Each AI pass is another chance for small changes to creep in. How far you can push an enlargement is covered in how to make a small image bigger without losing quality.

Faces: The Part That Matters Most

In a family photo, the faces are usually why the photo matters, and they are also where AI is most likely to go wrong. When a face is blurry or scratched, a restoration model draws in features that fit a typical face, which can subtly change:

• Eye shape and spacing
• Nose and mouth shape
• Jawline and face shape
• Hairline and eyebrows
• Distinctive marks such as moles or scars

After restoration, compare each face with the original scan. Ask someone who knew the person if possible. If a face looks sharper but less like them, use a gentler result or keep that area closer to the original. A slightly soft face that is clearly Grandma is better than a crisp face that isn't.

Restoration Is Not Colorization

Restoration repairs what is there: damage, fading, noise and softness. Colorization adds colour to a black-and-white photo, and those colours are a prediction. The AI doesn't know what colour a dress or a car really was; it picks something plausible.

Colorization can be a lovely way to bring a photo to life, but:

• Keep the restored black-and-white version alongside it.
• Don't present AI colours as historical fact.
• If someone remembers the real colours, their memory is more accurate than the AI.

The same goes for heavily faded colour prints. Colour balancing makes them look better, but the result is a best estimate, not a guaranteed match to the original film. If you want the opposite, giving a modern photo that faded film look, see how to turn your photo into a 1980s vintage photo.

Tears and Missing Pieces

When part of the print is physically gone, the software has nothing to recover. It generates content based on the surrounding area. This can work well for plain backgrounds, walls or sky. For people, objects or anything with meaning, treat the filled area as a reconstruction. If the photo has historical or family-record value, keep the damaged master scan archived with the restored version, so nobody later mistakes the generated part for the real one.

A Note on Grain and Softness

Old photos are naturally a little soft and grainy, and that is part of how they look. Pushing noise reduction and sharpening until the photo looks like it was taken on a modern phone usually makes it feel fake. Aim for a clean, readable image that still looks its age. If you want more control over these two steps, see how to fix a grainy photo online and how to fix a blurry photo.

For problems that apply to any low-quality image, such as compression from a forwarded copy, how to make a low-quality image HD explains how to diagnose them.

Final Check Before You Share or Print

• Normal size: does it look natural overall?
• Close up: any strange textures, smudges or warped edges in repaired areas?
• Faces: do they still look like the real people?
• Background: have objects, text or patterns changed?
• Versions: is the master scan still untouched and backed up?

Then export the restored version as a high-quality JPG for sharing or PNG for archiving, and keep it next to the master scan rather than replacing it.

Frequently Asked Questions

Can AI restore a badly damaged photo?

AI handles dust, small scratches, fading, noise and mild softness well. Large tears, missing corners and destroyed faces are harder, because the AI has to invent what was there. The result can look convincing but should be treated as a reconstruction.

What resolution should I scan an old photo at?

A common recommendation is 600 dpi for standard-size prints, and higher for very small prints you want to enlarge. Scanning at a higher resolution captures more of the real detail, which gives the AI a better starting point than upscaling a low-resolution scan later.

Will AI change my relative's face?

It can. When a face is blurry or damaged, the AI fills in plausible features that may not match the real person. Compare the eyes, nose, mouth and face shape with the original, and use a gentler restoration if the person looks different.

Is restoration the same as colorization?

No. Restoration repairs damage and improves clarity. Colorization adds colour to a black-and-white photo, and the colours are the AI's guess rather than the real ones. Keep the black-and-white restoration even if you also make a coloured version.

Is it safe to restore family photos online?

It depends on the tool. Some services upload your photos to their servers. Image EditPro's Old Photo Restoration tool processes the image in your browser, so the photo does not leave your device. Check how any other service handles files before using it.