ImageTextRemoverA PictureEditor.com tool

One operation, in depth

This site does one thing: it takes burned-in lettering out of a picture and paints the ground back across it. It does not crop, it does not add type, it does not read what the letters say. Everything it does happens in the tab you are reading this in.

The half that is finished

Working out what should stand where the letters were is a solved problem, and this site solves it two ways. The plain one relaxes the pixels inward from the border of your box with edge-aware weights, downloads nothing, and finishes before you have taken your finger off the button. The other is MI-GAN at 512 — 26,639,447 bytes, MIT-licensed, fetched once from the portfolio’s own object store when you press a control that names the size, and kept on your device afterwards with a delete button next to it.

The half that is not

The name of this site promises that it finds the lettering for you, and today it does not. You draw the rectangles. That is worth saying in the first paragraph of this page rather than in a footnote, because finding it out afterwards is worse.

The class of model that would do it is a scene-text detector — the kind that returns polygons around lettering without reading a word of it. Three things have to be true before one ships here: the licence has to be genuinely open, because these pages carry advertising and a research-only weight file is disqualified however well it performs; the file has to be small and quick enough to sit on top of the fill’s cost rather than double it; and it has to survive single-threaded WebAssembly, because cross-origin isolation breaks the advertising and multi-threading is therefore off the table permanently. No candidate has cleared all three, so there is no detector in this build and nothing on this site claims one.

Why the rectangles are not a stand-in

Because a caption bar, a date stamp and a screenshot header are rectangles already. Tracing a rectangle with anything rounder is busywork, and for these shapes a drag beats it outright — which is why the manual path is the product rather than a placeholder for the detector. When the detector does land, its proposals arrive in the same box list you already use, with a confidence dot, and nothing else about the flow changes.

Other things not in this build

The queue cannot pick up where it left off after a closed tab. There is no translucency estimator, so a translucent caption band is your judgement rather than the site’s. The joins between sections are feather-blended rather than gradient-matched, which is why a clean sky can still show a faint ripple. Two of the five planned pieces of reading are drafted and unpublished. All of that is written down here rather than discovered later.

Who made it

The same people behind PictureEditor, which is where the picture engine, the privacy posture and the performance floors come from. Everything above those foundations — every word, the palette, the mark, the box list and the arithmetic printed before a run — belongs to this domain and to nothing else in the family.