AI PHOTO REPAIR
Repair scratches and damaged areas in a photo.
Brush over one scratch, crease, tear, or missing patch at a time. The repair mode continues visible tone, grain, edges, and texture through the selected damage while keeping the rest of the photo available for close comparison.
STEP 1
Upload your photo
BEFORE / AFTER
Damaged photo repair: before and after
The divider moves on its own to reveal the before-and-after result. Drag it anytime to examine the local edit at your own pace.
Repair scratches without changing the people
This aligned example removes narrow scratches and creases from an old black-and-white photo while preserving the two people, fence, clothing, framing, and historical monochrome appearance.
Repair damaged photoHOW IT WORKS
How to repair a damaged photo in three steps
Scan or photograph it clearly
Use the highest-resolution copy available and keep an untouched original before making any generated repair.
Mark one defect at a time
Brush narrowly over a scratch, crease, dust mark, or missing patch without covering nearby facial features or intact edges.
Compare at full size
Check grain, tone, straight lines, facial identity, and the start and end of the repaired area before downloading.
DETAILED GUIDE
What AI photo repair can—and cannot—restore
Photo repair is a local reconstruction task. The model sees the visible pixels around a selected defect and generates a plausible continuation through the damaged area. It can often continue sky, clothing, walls, wood grain, film grain, and simple facial background, but it cannot recover an exact historical pixel that no longer exists in the source.
Narrow defects are the most controllable. Work from one scratch or crease to the next, leaving intact eyes, mouths, jewelry, writing, and architectural edges outside the mask. When a tear crosses several surfaces, split it where the background changes so each repair has one clear texture and lighting problem to solve.
This page does not automatically colorize, sharpen every face, remove all blur, or transform the whole photograph. Those are different restoration tasks. Its purpose is to repair selected damage while preserving the original composition, people, grayscale tone, and visible character of the source.
Use the aligned comparison after every pass. Look for softened faces, repeated grain, broken fence lines, invented clothing detail, and tonal patches. If a result changes identity or too much context, restart from the untouched scan with a smaller mask rather than repairing the generated output repeatedly.
Damage this repair workflow is designed for
Surface scratches
Remove narrow white or dark scratch lines when intact texture remains visible on both sides.
Fold marks and creases
Repair a thin paper crease in sections, especially where it crosses clothing, faces, or background geometry.
Dust and small spots
Clean isolated scan marks while protecting authentic film grain and nearby facial detail.
Small missing patches
Rebuild a limited damaged corner or background area when surrounding pixels provide enough reliable context.
When to use a damaged photo repair tool
Use this repair damaged photo workflow when the problem is local: a scratch, crease, tear, dust mark, or small missing region. It is especially useful when cropping would remove an important part of the photo and the defect is surrounded by visible texture that can guide reconstruction.
For full photo restoration involving colorization, face enhancement, deblurring, or exposure correction, use purpose-built tools and keep those steps separate. InpaintingAI focuses on the area you mark so you can review exactly what changed and preserve everything outside the selection.
PRACTICAL CHECKS
Damaged photo repair FAQ
- Keep an untouched high-resolution scan.
- Repair narrow connected defects separately.
- Protect eyes, mouths, writing, and unique details.
- Compare grain and tone at 100% zoom.
FAQ
Can AI recover the exact pixels lost under a scratch?
No. It generates a plausible continuation from visible context; it cannot retrieve image information that is absent from the scan.
Should I mark every scratch at once?
Work one defect or one connected section at a time. Smaller repairs are easier to inspect and retry without changing unrelated detail.
Does this page colorize or sharpen the whole photo?
No. It repairs selected damaged regions. Full-image colorization, deblurring, and global enhancement are outside this page's scope.
What if a crease crosses a face?
Use a narrow mask and inspect identity carefully. If important facial detail is missing, a generated repair should be treated as an approximation rather than a factual restoration.
RELATED IMAGE TOOLS