ByteScope

Redact Screenshot

Paste a screenshot, cover the private parts, copy it straight back — with no blur on offer, because blurred text can be reconstructed.

Your files never leave your browser — all processing is local.

Solid fill by default, and no blur at any setting

Gaussian blur can be undone. An attacker who knows the font — and on a screenshot they always do — renders candidate strings, blurs them the same way and compares, which is cheap enough for a card number or a six-digit code. There is no blur strength that fixes this, so this page does not offer blur at all. Pixelation is the same attack over a coarser grid, so it is offered only where the blocks are large enough to swallow whole characters, and warned about everywhere else. Solid fill is the default because its output does not depend on its input.

Drop a screenshot here, or press Ctrl+V

or click to browse — PNG, JPEG, WebP, GIF. Nothing is uploaded.

About this tool

Start with the reason this tool looks different from every other screenshot editor: blurring text does not destroy it. A blur is a fixed, known transformation, so an attacker who wants the card number under the smudge never needs to invert anything — they render candidate strings in the same font, blur them the same way, and keep the one that matches your pixels. For a 16-digit card number, a phone number or an email address the candidate space is small enough that this is practical, and it has been demonstrated in public more than once. Pixelation is the same attack with a coarser grid: every output block is just the mean of the source pixels beneath it, so it leaks the same information block by block. Both fail catastrophically over text — and text is what screenshots are made of. So this tool does not offer blur at all: no parameter makes it safe, so it cannot be offered even behind a warning. The default is a solid fill, which overwrites the pixels outright, and if you pick pixelation anyway, the tool warns when the block size is too small to be safe for the region it covers.

What the saved file actually contains

The export is a flattened redraw: the redaction is baked into a new bitmap, not stacked on top of the old one. The black box is not an object that a viewer, an undo command or a PDF editor can lift off the file — there is no layer, no mask, and no original hiding underneath, because the covered pixels were never written to the output at all. That distinction has burned real people: redactions applied as removable shapes in document editors have been peeled off court filings and leaked memos more than once. On the way out, EXIF is stripped as well, so the file carries no capture time, device model or any other tag the screenshot tool wrote into the original.

Paste in, copy out

The fastest path never touches the disk: press Ctrl+V (⌘V on a Mac) to paste a screenshot straight from the clipboard, drag rectangles over what has to go, and copy the redacted result back to the clipboard for the chat or ticket it is headed to. Drag-and-drop and a file picker are there for saved images, and the download button writes a PNG. Annotations — box, arrow, text label — are included because a redacted screenshot usually also needs a “look here”, but be clear about what they are: they mark the picture and they cover nothing. Only the solid fill and, block size permitting, the mosaic remove information, and the UI keeps the two groups visually apart so one is not mistaken for the other.

Nothing leaves the browser

The screenshot you most need to redact is exactly the screenshot you least want to upload somewhere first — the order matters. This page is a static file with no API behind it: the image is opened locally, the redraw runs in your tab, and the export is generated by your browser. There is no upload request to find in the network tab, and the page keeps working with the connection off. Nothing to install and no account, either.

Frequently asked questions

Why is there no blur option?

Because there is no safe way to offer one. Blur does not delete text — it transforms it in a fixed, reproducible way. An attacker renders candidate strings in a matching font, applies the same blur, and compares against your pixels; for structured secrets like card numbers, emails and phone numbers the candidate space is small and the attack is practical, not theoretical. A warning label would not fix that — people would still ship reversible redactions. Solid fill destroys the information. Blur only makes it hard for a human to read at a glance, and a human glancing was never the threat.

Can the black box be removed from the saved file?

No. The export is a flattened redraw: a new image is rasterised with the fill in place of the covered pixels, and the originals are never written to the output. There is no shape to select, no layer to hide, nothing under the box to recover — which is exactly the property document-editor redactions famously lack, where the “redaction” is a removable rectangle sitting over intact text. If you want to check, open the exported file in any editor: the black region is just black pixels.

How coarse does the mosaic have to be?

Coarser than feels natural. Each mosaic block is the average of the pixels it replaces, so blocks that are small relative to the glyphs underneath preserve enough structure to match candidate strings against. There is no universally safe number — it depends on the text size in your screenshot — which is why the tool compares the block size against the region you are covering and warns when the ratio is unsafe. When in doubt, use the solid fill. The mosaic exists for showing that something was there without showing what — not for protecting high-value secrets.

Is my screenshot uploaded?

No. The image is opened with the File API or read from the clipboard, the redraw runs in JavaScript in your tab, and the export is produced by your browser. The page is a static file with no server behind it, so there is nowhere to upload to even in principle — you can watch the network tab stay empty, or cut the connection and keep working. For a tool whose whole job is handling images you do not want seen, that is the property that matters most.

Is EXIF metadata removed too?

Yes. The export is a freshly encoded image, and no metadata from the original is carried across — capture time, device model, software tags and location all stay behind. This happens by construction rather than by a cleaning pass: the output is a redraw, so it only ever contains what the redraw wrote. The page also checks the original for you by actually walking its container — a PNG or WebP carrying EXIF is reported as such, and a container it cannot walk is reported as unchecked rather than assumed clean. If you want a second opinion, drop the exported file into the EXIF viewer and see for yourself.

Related tools