Upscale old photos with AI to print quality.

The first step in upscaling any old photo is getting the best possible digital version of the original. For physical prints, use a flatbed scanner at the highest DPI setting available — 1200 DPI is ideal for small prints, 600 DPI is fine for standard-size photos. The goal is to capture every bit of detail the physical print contains, so the AI has maximum information to work with when generating new detail at higher resolution. Even a tiny improvement in scan quality translates to a noticeably sharper upscaled result.

For old digital photos — images from early digital cameras, old phones, or social media downloads — upload the original file without resizing. Many people make the mistake of upscaling in a standard photo editor first, then uploading. This just creates a larger blurry file. The AI needs the actual original pixels to generate authentic new detail. If you only have a small thumbnail or compressed web image, that still works — the AI can add impressive detail — but the raw original always produces the best result.

Face-aware upscaling is where our tool really differentiates from generic upscalers. Old group photos, school class pictures, military unit photos, and team portraits often have faces that are only 20-50 pixels tall. At that size, standard upscaling produces smooth, featureless blobs. Our AI detects each face and applies a specialized facial super-resolution model that knows the structure of human faces. It reconstructs plausible eyes, noses, mouths, and expressions from surprisingly little data. For genealogy and family history, this capability is invaluable.

After upscaling, consider what you want to do with the result. For printing, order from a quality photo lab rather than a drugstore kiosk. Photo labs use better paper, inks, and color profiles. For sizes up to 8x10, any well-upscaled photo will look excellent. For very large prints (16x20 or bigger), start with the highest quality scan possible and consider running the upscale twice. For digital display on modern screens, a single upscale is usually sufficient because screens are more forgiving than print.

Combining upscaling with other AI tools produces the best results. The recommended workflow depends on the photo. For a faded, damaged old photo that is also small: restore first to fix damage, then upscale the restored version. For a clean but tiny photo: upscale directly. For a small black-and-white photo: restore, upscale, then colorize. Each step builds on the previous one, and the order matters because each tool works best on a clean input.

Batch upscaling is practical for old photo collections. If you have scanned a box of small prints and want them all enlarged for a family history book or album, upload the batch and process them together. The AI handles each image independently, so different sizes and subjects all get appropriate treatment. The Family plan at 150 credits is sized for exactly this kind of project.

A word about expectations: AI upscaling adds plausible detail, not the actual original detail that was never captured. A face upscaled from 30 pixels will look like a clear, natural face, but it will be the AI's best prediction of what that face looked like, not a forensic reconstruction. For family photos, this is perfectly fine — the result is a pleasing, recognizable image. For forensic or legal purposes, upscaled images should be clearly labeled as AI-enhanced.