Skip to content
ToolBoxGeniehome

Image Redaction Tool

Image Tools · Added

Mark the parts of a screenshot you do not want to share, and this overwrites the pixels underneath before writing the file — a box drawn over something is not the same as the thing being gone, and that difference is how redactions in documents get undone. It runs entirely in your browser, and it counts how many distinct colours survive in each covered area so the claim is measured rather than asserted.

Read and edited on your device, up to 20.00 MB. It is never uploaded.

How to use the image redaction tool

  1. 1Choose an image. It is read on your device and never uploaded.
  2. 2Drag a box over each area to cover, or add one and type its exact position.
  3. 3Pick how to cover it: pixelate, blur, or a solid fill — and set the strength.
  4. 4Redact, check the exported preview, and download the PNG.

Examples

An address on a screenshot

Input
A 340 × 40 px area, solid fill
Result
One colour left in that area — nothing to recover

A face in the background of a photo

Input
A 200 × 200 px area, blur at 25% strength
Result
Blurred and blended with its surroundings rather than boxed

Blur suits a face. It is the wrong choice for anything typed.

Pixelation that is too fine to trust

Input
A wide area at low strength, leaving many blocks across it
Result
A warning that pixelated text has been recovered by rendering candidate strings and comparing

About the image redaction tool

Covering is not removing

The distinction that matters is between hiding something visually and destroying it. In a word processor, a design tool or a PDF, a black rectangle is an object with coordinates and the text underneath is still text: it can be selected, copied, extracted by a script, or simply revealed by moving the rectangle. Several well-publicised leaks have come from exactly that, in filings and reports prepared by people who thought they had redacted them.

A raster image has no such structure. Once a pixel's value has been replaced there is no earlier value stored anywhere, and the exported file contains only what is drawn. That is why redacting a screenshot is more reliable than redacting the document it came from — and why exporting to an image is a reasonable last step when the document format cannot be trusted.

Choosing between the three modes

A solid fill is the only mode with nothing left to invert. It replaces the area with a single colour, and the colour count afterwards is one. If what is covered would matter — an account number, a name, an address, a key — this is the one to use, and its ugliness is the point: it looks like a redaction because it is one.

Pixelation is destructive when the blocks are large enough, and its appeal is that it reads as redaction while keeping the shape of the page. The risk is fine blocks over predictable content, where averaging leaves enough signal to match candidates against. Blur is for photographic content — faces, plates, screens in the background — where it blends with its surroundings and the underlying content is not from a small enumerable set.

What to check before sharing

Cover the obvious thing first, then look at the rest of the frame. Screenshots leak from the edges: a browser tab title, an autocomplete dropdown, a notification, the taskbar, a second monitor caught in the same capture, an email address in a signature further down. Reflections in glasses and screens have identified people in published photographs more than once.

Then consider what the image is not showing but still contains. A photograph taken on a phone usually carries GPS coordinates and a timestamp in its metadata; re-encoding here removes them, but the original you keep still has them. And if the covered information appears anywhere else in the same picture — the same account number in a second column, the same name in a header — covering one instance achieves nothing.

Frequently asked questions

Is the covered part really gone?
In the exported file, yes. The pixels are overwritten in the image data before it is encoded, so there is no layer to move aside and nothing underneath. The tool counts the distinct colours left inside each area from the exported pixels and shows the number — a solid fill leaves one. What it cannot do is remove information that is still visible elsewhere in the picture: a reflection, a second copy of the same document, a name in a window title bar.
Why is a black rectangle in a PDF not the same thing?
Because a PDF is a document, not a picture. A rectangle drawn over text is an object sitting on top of text that is still in the file, and selecting, copying or opening it in a different reader recovers what it was meant to hide. That has cost governments and law firms real documents. An image has no layers — once the pixels are overwritten and the file is written, there is nothing to recover.
Can pixelation be reversed?
Sometimes, and it is worth knowing when. Pixelation averages each block, and where the underlying content is drawn from a small known set — digits, a typeface at a known size — an attacker can render every candidate, pixelate each one and compare. Published work has recovered pixelated text this way. Large blocks over a short string leave too little to match against, which is why the strength setting exists and why the tool warns when the blocks are small relative to the area. For anything that must not leak, use a solid fill.
Is blur safe?
Less so than pixelation for text, and it is offered for a different job. A blur is a linear operation and undoing one is a studied problem, so it is the wrong tool for an account number. It is the right tool for a face or a car in the background of a photograph, where the content is not from a small known set and a hard-edged box would look worse than the thing it hides.
Does it remove the photo's metadata too?
Yes, as a side effect of how it works. The image is redrawn on a canvas and re-encoded as a PNG, and none of the original file's EXIF — camera, timestamp, GPS coordinates — survives that. If metadata is the only thing you want to see or strip, the image metadata viewer shows you what a file is carrying.

Related tools