Screenshot redaction tool
Drop in a screenshot, drag boxes over the parts nobody else should see, and download a flattened copy. The download is a fresh image built from the canvas, so the covered pixels are genuinely gone from it — there is no layer to move and nothing underneath to recover. The file on your disk is never altered.
Pixelation is not redaction. A mosaic keeps the original colours in an averaged form, and text is highly predictable, so pixelated words have repeatedly been recovered by matching the pattern against every candidate rendered the same way. Never use it for a password, an account number, a postcode or an address. A solid box destroys the pixels outright and cannot be undone.
Add a box by number
Dragging needs a pointer, so the same box can be entered here as four numbers in image pixels and added with the keyboard.
The image is decoded and redrawn inside this browser tab. It is not uploaded, and no copy is kept anywhere — close the tab and both the screenshot and the boxes are gone.
Why a flattened image is the only safe redaction
Most redactions that end up in the news failed the same way: the covering was a separate object drawn over the original, and the original was still in the file. A black rectangle in a PDF sits above the text layer, and the text underneath can be selected, copied or extracted with a command-line tool. A shape in a design file is a layer someone can hide. Even a screenshot of a page where an element was set to a low opacity keeps enough signal to be pulled back out. In every case the information was never removed; it was merely hidden from a casual viewer.
This page works the other way round. The screenshot is decoded into pixels on a canvas, your boxes are painted directly onto those pixels, and the download is encoded from the canvas as it stands. There is no list of shapes in the output and no original layer beneath them — the bytes that described the covered area have been overwritten with the bytes for a solid colour. Nothing in the file records that a box was ever drawn, which is exactly the point.
Two useful consequences follow. The first is that your original file is untouched: everything here happens on a copy held in memory, so if you get a box wrong you still have the source. The second is that because the download is re-encoded from raw pixels, every scrap of metadata in the original is dropped along the way — no Exif, no XMP, no timestamps. If that is all you need, and you would rather keep the image itself bit-for-bit identical, use the EXIF metadata remover, which removes the metadata without re-encoding a single pixel.
What to cover, and what people forget
The obvious targets are names, email addresses, phone numbers, account and card numbers, home addresses and anything that looks like a token or a key. The ones that get missed sit around the edges of the screenshot: the browser tab title, the URL bar, an autocomplete dropdown, the notification that slid in while you were capturing, the profile photo and account name in the top right corner, the taskbar, the file path in a window title, and the other open tabs.
Identifiers that look harmless are worth a second thought too. A customer or order number is meaningless on its own but joins two screenshots together, and an internal hostname or ticket reference tells an outsider more about a system than most people expect. If you are posting a bug report publicly, it is usually quicker to cover a whole region generously than to place four small boxes precisely.
Finally, check the shape of what remains. A box sized exactly to a word still leaks the length of that word, and a series of boxes of obviously different widths can narrow down a list of candidates. Where it matters, cover a bit more than you need and keep the boxes uniform.
Questions people ask
Can the black boxes be removed from the downloaded file?
No. The downloaded file is a flat raster produced by the canvas, so a covered area contains nothing but the solid colour that was painted over it. There is no hidden layer, no history and no original copy inside the file to recover. This is different from a PDF or a design document, where a covering shape is an object stacked on top of content that is still present, and it is why converting a redacted document to a flat image is a standard last step in serious workflows.
Is pixelating or blurring good enough?
Not for anything you actually need hidden. Both operations average nearby pixels rather than discarding them, so the result still carries information about what was there. Text is the worst case, because it is drawn from a small, known set of shapes in a predictable font: an attacker can render every candidate string, apply the same mosaic, and compare. Published research and practical tools have recovered pixelated numbers and short strings this way. Use a solid box for passwords, account numbers, addresses and anything else that would matter if it came back. Pixelation is fine for making a face less recognisable in a casual screenshot, and nothing stronger than that.
Can I redact without a mouse or a trackpad?
Yes. Dragging is not reachable from the keyboard, so there is a set of number fields under the canvas for left, top, width and height, measured in the image's own pixels, and an Add rectangle button beside them. Tab to the fields, type the four numbers, activate the button, and the box appears in the list with the same coordinates a drag would have produced. Each box in the list has its own Remove button, and Undo takes off the most recent one.
Does the downloaded image still contain the original metadata?
No. Because the file is encoded fresh from the canvas, it starts from pixels alone — Exif, GPS coordinates, XMP, IPTC and the original timestamps are all absent from the download. That is a genuine side benefit of this approach, though it comes at the cost of a re-encode. If you want the metadata gone but the image untouched, the EXIF remover linked above rebuilds the container without decoding the pixels at all.
Should I download as PNG or JPEG?
PNG for anything with text in it, which is most screenshots. JPEG compression works by discarding fine detail, and the finest detail in a screenshot is the edge of every letter, so text picks up a faint coloured haze and looks slightly dirty. PNG is lossless and keeps the characters crisp, at the cost of a larger file. Choose JPEG only when the screenshot is mostly photographic, or when a size limit forces your hand — and note that the redaction is equally permanent either way.
Does the size of the screenshot matter?
Yes. This page accepts images up to 40 megapixels; larger images are refused. Within that limit, the canvas holds the image at its full resolution and is displayed scaled down to fit the column, with the drag coordinates converted back to real pixels, so the boxes land where you put them. Very large images use more memory and take a moment longer to encode on download. If a phone tab runs out of memory, scaling the image down first and then redacting is the practical fix.