About this page and the decisions behind it

This is one page that does one thing: it finds the dust, specks and hairline scratches on a scan of a damaged print, lets you correct what it found, closes them from the picture around them, and pulls the range back out of a photograph that has faded. It is not a photo editor with a repair feature. There is no account, no gallery, no project file and nothing to subscribe to.

Three decisions worth explaining

The mask comes before the repair

Most tools in this category go straight from a file to a result, and the result is a photograph that has been altered in ways nobody can enumerate. The single irreversible judgement in restoration is deciding which pixels are damage, and that judgement depends entirely on what the photograph is of — a detector cannot know that a thin dark line is a mooring rope rather than a scratch. So the marks are drawn first, over an untouched picture, and two brushes let you argue with them. It is slower. It is the only version of this that can be checked.

The work happens on your machine

Photographs that arrive here are usually the only copy of something. Sending them to a server would mean asking for that trust, holding the file somewhere, and being a party to it in a way that a page which does the arithmetic locally simply is not. The detection, the fill and the tone curves are written in TypeScript and run in a background thread in your browser. There is no upload step because there is nothing to upload to.

It refuses a good deal of work

Silvering and mould are named as unrecoverable, colour is never added to a monochrome print, and the local fill says plainly that it cannot close a tear. There is now a model that will draw into one, and it is offered the same way: behind a control that names its size, described as invention rather than recovery, and off until somebody asks. Each of those would be easy to fake convincingly enough to satisfy a first glance. A page about damaged photographs that quietly invented a face and called it a repair would be worse than no page at all, so the refusals — and the one thing that is a guess — are written down where somebody can read them before they spend an evening.

What it costs, and who is meant to pay

Nothing to you, and advertising in the end. There is none of it today: Google AdSense has not approved this domain, so every placement is switched off at the source and no band, no label and no third-party request exists anywhere on the site. What is on the screen is the whole page.

When that approval arrives, the placements already written into the code start rendering, and they were positioned once, carefully: below the working surface only, never over the tool and never between the controls, each holding its height from the first paint so nothing shifts under a cursor when an advertisement lands. None of it reaches the photograph. The file is decoded and repaired inside this tab, an ad frame is a separate document, and there is nothing of yours on the far side of that wall for it to read.