Restore an old photo: repair scratches, stains, creases and fading, keep the original toning and faces, then compare every repair with the scan.
Restore an old photo: repair scratches, stains, creases and fading, keep the original toning and faces, then compare every repair with the scan.
One damaged scan and the restored image generated from it, in the supported ratio closest to your scan.
The recipe asks the model to keep the original toning, composition, framing, people and faces. Compare each face with the original scan first, then check that nobody moved, changed pose or lost an accessory.
Scratches, stains, creases and fading are repaired, and clarity and contrast are gently improved. Where damage hid detail, the repair is the model’s interpretation. Compare repaired details with sources before you accept them, because a generated restoration attempt can look sharper without recovering anything the scan lost.
Review the facial details where damage crossed an eye or mouth, and compare the damaged areas before and after repair. Keep uncertain facial details marked as uncertain. Treat reconstructed signs or handwriting as uncertain, because lettering can look readable and still be wrong. Keep the original scan beside the result.
Three common reasons to restore an old photo. In each one the restored image is a generated repair, and the original scan stays as the record.
Input ratio · Portrait
Start with a clear scan of the damaged source photograph and note scratches that cross faces or clothing. After generating, compare expressions and proportions before judging how clean the image looks. A removed scratch only helps if the person underneath is unchanged. Keep the original scan unchanged beside the restored version, and label the edit as a restoration.
Input ratio · Faded
Use the best digital copy you have and see whether the subjects are easier to tell apart. Compare shadows and highlights as well as faces, since stronger contrast can hide detail in dark clothing or pale backgrounds. Choose the result that serves the display, without claiming its tones are exact.
Input ratio · Group
Check every face, including people at the edges and small, shadowed ones. Prepare a family archive comparison with the original scan, the restored image and notes on uncertain areas. Ask relatives to help confirm who is pictured, and keep names and dates in separate records.
Upload your scan, check the fixed output, then compare faces and damaged areas before you download.
Add one JPG, PNG or WebP image up to 20 MB. Scan the whole print with its edges and keep the original scan unchanged as a separate file; a fresh scan holds more detail than a copy forwarded through several messages.
There is no style, ratio or size selector; the panel shows the fixed output, one PNG at HD from Nano Banana Pro in the ratio closest to your scan. The original toning stays, and adding color is a separate task.
Each old photo restoration result costs 45 credits, shown on Generate before you confirm. When the restoration is done, compare faces first, then each damaged area against the scan, and download it only if it serves the intended use.
AI photo restoration that repairs the damage and keeps the photograph itself the same.
01
Old photo restoration starts from your scan, so detail that never made it into the file cannot be brought back. Scan the whole print with its edges and any writing, and keep the original scan unchanged as a separate file.
02
The repair covers scratches, stains, creases and fading, then gently lifts clarity and contrast. Toning, composition, framing, people and faces are meant to stay as they were, so the result is still recognizably the same photograph.
03
A repaired face or line of handwriting can look convincing and still be wrong. Compare each face with the original scan first, treat reconstructed signs or captions as uncertain, and keep the damaged original next to the restored version.
Next steps for a restored scan.
Old photo restoration repairs damage and fading in an old photograph. This AI photo restoration template generates a repair of scratches, stains, creases and weak contrast while asking to keep the original toning, people and faces. Details that the damage hid are reconstructed, so the result does not prove the original face or scene has been recovered exactly.
No. Restoration addresses damage and faded detail; colorization adds plausible colors to a monochrome source. This recipe asks to preserve the original toning. Use Colorize Photo for a separate color concept, and keep the stages distinct rather than treating invented color as historical evidence.
Use the best available scan of the whole print, with its edges and any writing visible, and do not add filters that hide damage. A fresh scan holds more detail than a small image forwarded through several messages, and enlarging a file does not recreate detail lost to damage or compression. JPG, PNG and WebP are accepted, up to 20 MB, one image per run. Keep the untouched scan as a separate file.
It may generate a plausible face detail, but it cannot verify information absent from the source. Compare recognizable features with other reliable records or photographs where available. If a damaged eye or mouth is critical, a cleaner source scan is stronger evidence than a more polished generated guess.
No. A damaged sign, label or handwritten note can look convincingly repaired while containing invented letters, and the old photo restoration template is not a transcription tool for faded writing. Treat any reconstructed text as uncertain unless the original or another reliable record confirms it, and keep dates and names in separately verified captions.
Yes. Preserve the best scan unchanged and save the restored result separately. This supports comparison, future work from a better source, and an honest record of what was repaired. Keep historical names or dates in separately verified captions rather than expecting the generated image to establish them.
Not in the same run; this page only repairs damage and keeps the original toning. After restoring, use Colorize Photo to propose plausible color, Image Upscaler to add pixels for a larger print, or the AI photo editor to describe one further correction in words.
Yes. It repairs damage on both and keeps the photo's existing toning: a sepia print stays sepia and a faded color print keeps its colors rather than gaining new ones. To add color to a black and white photo, restore it here first, then use Colorize Photo on the repaired copy.
45 credits per run for one PNG at HD in the supported ratio closest to your scan, made with Nano Banana Pro. The quote appears on the Generate button before anything is charged.
Upload your best scan, review the credit estimate, and compare the restored image with the original before using it.