AI photo cleanup · launching soon
Remove dust, scratches, stains and unwanted objects from any photo, and bring damaged family prints back without the hours of clone-stamping.
How it works
Drop in a phone picture, a camera file or a scan of an old print. Your original stays untouched; DirtAI works on a copy.
Paint over dust specks, scratches, stains, a stray tourist or a power line. Use a small brush for detail, a big one for blotches.
The model reconstructs what was underneath from the surrounding texture, grain and light. Compare before and after, refine, export.
What it cleans
From a speck on the scanner glass to a stranger in the frame, the same brush handles it.
Clears the pepper of white and black dots that scanners and dirty sensors leave behind, without smearing fine grain.
Long hairline scratches, fold lines and cracked emulsion are filled along their length so the repair follows the image, not a blur.
Water marks, tape residue, foxing and coffee rings are toned back so faces and skies read cleanly again.
Brush out photobombers, bins, signs or cables. The background is rebuilt to match perspective and lighting.
Tidy cluttered backdrops behind portraits and product shots without cutting out or replacing the scene.
Every edit is a layer on a copy. Toggle before and after, undo any stroke, and keep the original file as it was.
Use cases
An honest note
Early access
We're building the brush, the models and the non-destructive editor now. Join the waitlist and we'll email you when it's ready at dirtai.xyz.
Until then, get your archive ready: start with the scanning guide and back your prints up, so they're safe and waiting.
FAQ
Not yet. DirtAI is in development and launching soon. Join the waitlist on this page and we'll email you when it's ready. In the meantime, our guides cover scanning, manual cleanup and archiving.
No. DirtAI is designed to work on a copy, with every brush stroke kept as a reversible edit. Your original file stays exactly as you uploaded it.
It can fill missing areas with plausible detail drawn from the surrounding image, which works well for skies, walls, grass and fabric. For a missing eye or a torn-off face, any fill is a reconstruction, not a recovery, and we'll label it that way.
For most prints, 600 DPI is a good default; use 1200 DPI for small prints you might enlarge. Our scanning guide walks through DPI, cleaning and file formats in detail.
Cleanup and repair come first. Colorization is a different problem: AI guesses colors that were never recorded. We've written about what it can and can't do before deciding how to offer it.
The product isn't live, so no photos are collected today; the only thing this site stores is what you enter in the waitlist form. Before anyone uploads a single photo, we'll update our privacy policy to cover photo storage and deletion.
From the guides
A practical plan for keeping family photos safe for decades: storing physical prints properly, organizing digital scans, and following the 3-2-1 backup rule.
How AI colorization works, why its colors are educated guesses, where it goes wrong, and how to colorize family and historical photos accurately and ethically.
Find sensor dust spots in your photos, remove them across a whole shoot in a few minutes, and clean your camera sensor safely so they don't come back.