Blur or Pixelate Part of an Image
Drag a rectangle over anything in a screenshot that should not be readable — a name, an email address, an API key, a house number in a listing photo — and choose how thoroughly it goes: pixelated, blurred, or replaced with a flat block. The saved file has those pixels overwritten, not hidden behind something, which is the difference between a redaction and a rectangle someone can delete. The image is decoded and redacted inside this page, so the screenshot you need to censor is never uploaded in order to censor it.
Drop a screenshot to redact
One file at a time, up to 50.0 MB. Everything happens in this tab — the screenshot you are redacting is never sent anywhere.
Nothing loaded yet. Drop a screenshot above — it is decoded and redacted inside this browser tab, which is the point: a screenshot you need to censor is exactly the sort of file that should not be uploaded to somebody else’s server to censor it.
Annotate as well as redact
Arrows, callouts, and a caption around the part you did not black out — on a Moda canvas.
Try Moda free →Why a black rectangle in a PDF is not a redaction and this is
The famous failure mode is a shape drawn over text in a document editor: the shape is a separate object, the text underneath is still in the file, and anyone who selects the region or opens the file in another program reads it straight off. That has leaked witness names, salaries, and unredacted court filings more than once. This tool does not have that failure available to it: there are no layers and no objects. The image is drawn onto a canvas, the selected areas are painted over on that canvas, and the file is encoded from the result. By the time anything is saved, the pixels that were under a region are gone from what is being encoded — there is no second copy for the file to carry.
Solid, pixelate, blur — in that order of strength
A solid block replaces every pixel in the area with one flat colour, so there is provably nothing left: it is the right choice for a password, a key, or an account number. Pixelation samples each block down to a single colour and blows it back up, so what survives is one value per block — recoverable only in the sense that a very weak setting over a short, predictable string leaves a pattern. Blur is the weakest: it is a convolution, and the published attacks on redacted screenshots have all been against light blurs over short strings with a small alphabet, like a six-digit number. Blur looks the most natural and is fine for a face in the background of a photo. It is the wrong tool for a credential.
Strength scales with the area, not with the file
The strength control means "how coarse, relative to this region" rather than a fixed number of pixels. That is deliberate: a twelve-pixel mosaic that thoroughly hides an email address on a 1200-pixel screenshot leaves the same address perfectly legible on a 6000-pixel one, and a redaction that only sometimes redacts is worse than none, because by then it has been published. Here, the same setting produces the same visual result whatever you drop in, and the preview you are looking at is rendered by the same code that writes the file — at a different scale, but from one definition of the effect.
What else leaves with the pixels
The saved file is re-encoded from a canvas, which has a side effect worth knowing about: the original file’s metadata does not survive. EXIF, including the GPS coordinates a phone writes into a photo, is not carried into the output. That is usually exactly what you want on this page — there is little point blurring a house number in a listing photo that still names the street in its metadata. If metadata is the whole of your problem and the pixels are fine, the EXIF viewer strips it without touching the image. And if the thing you need gone is at the edge of the frame, cropping it away is stronger than covering it.
Frequently asked questions
Can the blurred or pixelated part be recovered from the saved file?
Not from the file itself — it contains only the blurred or pixelated result, with no copy of the original underneath. What has been demonstrated is different: reconstructing a *short, predictable* string, such as a number with known formatting, by generating candidates, applying the same effect, and matching. That attack needs a weak setting and a small set of possibilities. It does not apply to a solid block at all, and a strong pixelation leaves too little to match against. For anything that must never be readable, use the solid block.
Is my screenshot uploaded?
No. It is decoded into a bitmap in this page, redacted on a canvas here, and encoded here. No request carrying the image is made — which is the whole reason this exists as a browser tool: a screenshot that needs censoring is the last file you should hand to a server you do not control in order to censor it.
Does the redacted file stay the same size and quality?
The pixel dimensions are unchanged — a 2560×1440 screenshot saves at 2560×1440. Quality depends on the format you pick. PNG is lossless and the right default for a screenshot, because text and UI edges are exactly what JPG compression smears. JPG and WEBP recompress the whole image, which is fine for a photo and visibly worse on a screenshot of text.
Can I redact several areas at once?
Yes — drag as many rectangles as you need, up to twenty on one image, and each is listed underneath so you can select, move, resize, or remove it. They all use the same effect and strength: mixing a blur and a solid block in one screenshot mostly signals which area you cared about most, which is not usually the signal you want to send.
Why can I not undo a redaction after saving?
Because that would mean the original pixels were still in the file, which is the exact failure this tool exists to avoid. Inside the page you can move or delete a region freely — nothing is committed until you save, and the file you dropped in is never modified, so your original is untouched on disk. The download is a new file, and it is the one that has nothing left in it.