Somewhere in your family there is a shoebox, and in it there's one photo that matters — a grandparent's wedding, the old house, someone who's gone. To restore old photos with AI well, you need to understand one thing before touching any tool: restoration is partly recovery and partly invention, and with the faces of people you love, the line between the two matters more than anywhere else in AI imaging. Our free AI image upscaler handles the recovery half. This post is the full craft.
Everything downstream is limited by the scan. Phone-photographing a print adds glare, lens distortion, and your ceiling light to a damaged image. Do this instead:
Upscaling models are trained on millions of photos, so they know what photographic grain, fabric, hair, and skin look like at higher resolution. Blur from a small or degraded print, softness from old lenses, grain amplified by time — these respond beautifully. A 2-inch faded print can come back as a sharp 8x10. Mild fading and color casts also correct well, because the information is still in the image, just compressed into a narrow tonal range.
Where the print is genuinely gone — a crease through a face, a torn corner, silverfish damage — the model doesn't recover, it guesses. For a background wall, a guess is fine; patch damage with the inpainting tool and nobody will ever know. For a face, be careful: aggressive "face enhancement" settings can quietly average your grandmother toward a generic face. The photo gets sharper and less her at the same time. The tell: compare eyes, nose bridge, and smile lines against the original at the same size. If a relative who knew them says "something's off," believe the relative over the algorithm.
1) Scan high. 2) Upscale 2x first, not 4x — inspect before going further; each step doubles both detail and invention. 3) Repair physical damage with inpainting, smallest masks possible. 4) Color-correct last, gently — a hundred-year-old photo that looks shot yesterday reads as fake; leave a little warmth and grain. 5) Export PNG, and store the original scan beside every edited version, labeled. You're the archivist now; future family will want the untouched file.
Colorization is pure invention — the model guesses that the dress was blue. It can be a beautiful gift, and it can also overwrite family memory ("no, that dress was green, I was there"). Best practice: present colorized versions as interpretations, always alongside the original. Never let the colorized copy become the only copy.
Photo restoration services charge $30–80 per image, and the shoebox has two hundred. Our upscaler is free with no per-image limit because it's one of the hundred doors into QADIR OS — where the entire pipeline runs locally on your machine. For family archives that's not a convenience feature, it's the point: your grandmother's wedding photos don't belong on anyone's server.
QADIR OS — the sovereign agentic operating system. 100 tools in your hands, your AI partner runs the loop.
Join the Waiting List