Use case
AI photo restoration for old and damaged photos
Photo restoration used to mean mailing a print to a specialist and waiting a week. Upload a scan, tick what is wrong with it, and get a restored copy back in about ten seconds.
Scratches, creases and fading repaired
Step by step
How to do it
Scan or photograph the original
Get the flattest, best-lit capture you can, with no glare. JPEG, PNG or HEIC up to 10 MB. Photo restoration quality tracks input quality more than anything else, so it is worth a second attempt at the scan.
Tick the damage to repair
Scratches and dust, tears and creases, stains, fading, softness, or any mix. Keep the original black-and-white or sepia tone, or have it colorized in the same pass. Point at the worst spot in the optional box, like “the face on the left is badly faded”.
Download, print, and share
You get back a clean copy to save, reprint, or send to family. Choose 2K for prints. If the first pass misses something, run it again.
Scanning the original
Restoration starts with the capture. A flatbed scanner at 600 dpi is a good setting for ordinary prints. Go higher, 1200 dpi, for small wallet prints and contact sheets, so the face has enough pixels to work with. Clean the scanner glass first, because dust on the glass looks like dust on the photo.
Uploads can be JPEG, PNG, WebP or HEIC up to 10 MB, so save the scan as a high-quality JPEG rather than a TIFF. A 600 dpi scan of a 4×6 print saved that way fits comfortably under the limit.
No scanner? Lay the print flat by a window in daylight, turn the flash off, and hold the phone directly above it so the edges stay square. If a print is stuck to the glass of an old album page, photograph it through the glass rather than peeling it off. Prying a stuck print loose can tear the image.
Picking what to repair
Tick only the damage you can see. Scratches & dust plus Fading, the default, suits most shoebox prints. Add Tears & creases when the photo has been folded or a corner is missing, and Stains for water marks or tape residue. Softness sharpens detail that has gone soft. It is the setting most likely to change a face, so leave it off unless the photo really needs it.
Use the Anything else? box to point at the worst spot: “the crease runs across the man’s face” or “the left side is badly faded”. Leave Color on Keep original tone to keep black-and-white or sepia. For a badly damaged photo, restore it first and then colorize the clean copy with Colorize B&W. That is two credits, and usually better than doing both in one pass.
What AI restoration can and cannot do
Scratches, dust, creases, stains and fading come out reliably, because the AI can see the photo on either side of the damage. Badly damaged faces are different. When most of a face is missing or blurred away, the AI reconstructs a plausible face rather than recovering the real one, and the result can look like a relative instead of the person. Check every face against the original and ask someone who knew them.
Keep the original scan. A restoration is a new image, not a replacement for the record. A professional restoration service typically charges $25 to $100 per image and takes several days. It is still worth it for an heirloom where a face has to be right. For the rest of the box, one credit per photo lets you restore them all and try again where a pass misses.
In the tool
Popular settings
Pick these in the tool. No prompt to write.
Scratches & dust + Fading
The everyday shoebox print.
Tears & creases + Stains
Folded, torn or water-damaged photos.
All damage · Also colorize
Repair and color in one step.
Softness · Keep original tone
Sharpen faces without changing the look of the era.
Questions about ai photo restoration: restore old and damaged photos
How much does photo restoration cost?
One credit per restoration at HD, two at 2K. Credit packs start at $4.99 for 10 credits. A professional photo restoration service typically charges $25 to $100 per image and takes several days.
What if part of the photo is missing or torn away?
The AI reconstructs small missing areas, like a torn corner or a crease across a face, by inferring from the surrounding image. Very large gaps are harder, and a reconstructed region is a plausible guess rather than a recovery of what was actually there.
Does this work on prints, not just digital scans?
Yes. Photograph the print with your phone, laid flat with no shadows or glare, and upload that. A flatbed scan gives a better restoration, but a careful phone photo works well.
Can it restore faces that have gone soft or blurry?
Usually. Tick Softness. Bear in mind that the AI is reconstructing detail rather than recovering it, so a heavily degraded face may come back subtly different.
