No. 07 — AI Tools · Cloud HDPro · on our GPU

Old photo, enhanced, not faked.

“Old photo” covers a scanned print from 1962 and a 0.3-megapixel JPEG from 2004, and the two want almost opposite treatment. Cloud HD handles both — this page is about deciding which one is on your screen before you upload it.

Pro · Cloud HDDeleted on completionUp to 25 MB · 24 MP
Enhance an old photoCloud HD · our GPU

How it works

  1. Pick the scan or the old digital file — JPEG, PNG or WebP up to 25 MB and 24 megapixels.
  2. It is uploaded to our server for that run and deleted when the run finishes, and in all cases within 24 hours.
  3. Download it at the same size, cleaned up with faces restored and faded colour lifted. One general Cloud HD run.

This page is the AI photo enhancer, tuned to old photographs and the order the work belongs in.

All Image Tools

AI Tools

Solutions by use case

Enhance an Old PhotoPro, on our server

Two different photographs answer to the word old. One is chemistry: a print or a negative whose dyes have shifted towards orange, whose silver has flattened towards grey, and whose detail was always limited by the stock and the enlarger. The other is early digital: a small sensor, an aggressive in-camera JPEG encoder and a memory card that cost real money, producing a file that is technically clean but starved of both pixels and bit depth. The Cloud HD enhancer addresses the compression damage and softness in either case and lifts a faded print towards its colour, at the same pixel size it was given.

Enhancement, restoration and colour are three tools with a fixed order. If people are the point of the photograph, the photo restorer is the stronger first move, because it is built around faces and refuses a result that has drifted from the person in your original. If the photograph is a street, a house or a crowd, start here instead — no face is required. Colour comes after either of them: colourising a photo works better on a cleaned image, and it produces a plausible reading rather than a verified record of what the room actually looked like.

Physical damage is explicitly out of scope, and it should be. A tear, a water stain, a bite taken out of a corner, silverfish tracks — none of these are touched by enhancement, and the result hands them back unchanged. Only the object remover fills an area back in, by inpainting under a brush mask you paint yourself, and it works best on small-to-medium regions with predictable surroundings such as sky, wall or fabric. It draws the HD sub-quota — 10 of your monthly runs may be HD-class — so it is worth saving for damage that genuinely matters.

Size comes last, and privacy is not an afterthought. An enhanced file at the original dimensions is the input the image upscaler wants, because an upscaler magnifies faithfully — including anything you have not fixed yet. As for the upload: nearly every MiniPx tool works entirely in your browser, and Cloud HD is the stated exception, a Pro-only opt-in path where your photo is uploaded to our server for that run and deleted when the run finishes, and in all cases within 24 hours. It is processed in the United States, encrypted in transit, and never used to train anything. Every result names the model that made it, currently RestoreFormer++.

How it works

  1. Work out which kind of old you have: A scanned chemical print and a low-resolution digital JPEG are different problems. Fading and softness against compression damage and missing pixels.
  2. Digitise properly if it is a print: Scan at 600 DPI, or photograph square-on in even daylight without flash, and crop away the background before uploading.
  3. Run the Cloud HD enhancer: JPEG, PNG or WebP up to 25 MB and 24 megapixels. Clean-up over the whole frame, a face pass where faces exist, faded colour lifted.
  4. Add colour or size afterwards, in that order: Colourise the enhanced result if it is monochrome, then enlarge last. Keep the untouched original beside every derivative.

Frequently asked questions

Is enhancing an old photo the same as restoring it?
They overlap and are not identical. Enhancement runs the whole photograph through a clean-up pass and restores faces where it finds them, and it never requires a face to be present. Restoration on the dedicated page is aimed squarely at people: it insists on finding a face and checks the rebuilt one against your original. Landscapes, buildings and group scenes belong here.
My old photo is a 640 by 480 JPEG from a 2004 camera. What actually helps?
Early digital files carry heavy compression damage — blocky flat areas and ringing round edges — which the clean-up pass handles well. What they also carry is a shortage of pixels, and enhancement returns the size it was given. Enhance first to remove the artefacts, then enlarge with the upscaler so it magnifies a clean picture rather than a damaged one.
Will it repair the tear across the corner of my print?
No. This tool does not repair physical damage: tears, stains, scratches and missing areas come back exactly as they arrived. That restraint is deliberate. For a small, well-defined blemish on a predictable background you can paint a mask over it with the object remover, which fills the area back in — but that is a different tool, and it draws the HD sub-quota.
Will the photo stop looking old?
Some of the period character comes from fading and softness and will genuinely lift. Some of it — the framing, the clothes, the tonality of the stock — is the photograph, and nothing here touches it. The intent is a clearer version of the same picture, not a modern re-shoot of it. Keep the untouched scan so both versions exist.
The photo is black and white. Should I add colour as well?
That is a separate tool and a separate decision. Colourisation produces a plausible reading of the scene, not a verified one — the model does not know what colour the dress was. If you want it, enhance first so the colour model has cleaner detail to work from, and keep the monochrome original as the record.