How to blur a face or license plate in a photo

A face, a number plate, an address on an envelope, none of it belongs in a photo you're about to post. Here's how to cover it properly, and why the usual soft blur isn't the privacy fix most people assume it is.

What redacting a photo actually means

Redacting means removing information from a photo so it can't be recovered, not just visually hiding it. Draw a box over a face or a plate, and that box needs to destroy what was underneath it well enough that nobody looking at the result, or running any tool over it, can get the original detail back.

That distinction, hidden versus actually destroyed, is the entire point of this guide, because it's exactly where a lot of well-meaning redaction quietly fails.

Why this tool doesn't use a soft blur

A typical Gaussian or CSS-style blur, the kind most photo editors apply with a slider, doesn't remove the underlying detail, it smears it. Smearing keeps most of the original signal intact, just spread across neighbouring pixels, which is a fundamentally different thing to genuinely deleting it.

That difference isn't theoretical. Deblurring a softly blurred licence plate or face back into something legible is a real, published, and increasingly automated technique. A blur that looks private to the eye can still hand back readable information to the right software, which makes it a false sense of security rather than an actual fix.

So this tool doesn't offer a blur mode at all. It offers two genuinely destructive options instead, solid fill and pixelation, both of which throw away far more information than a blur kernel ever does, and neither of which has a known way back to the original.

Fill versus pixelate: which to use

Solid fill draws a plain opaque block over the region, black by default. It's true redaction in the strictest sense, every pixel underneath is completely gone, replaced with a single flat colour. Use it whenever you don't need the covered area to still read as 'a face was here', a document number, an account field, anything where the content itself needs to disappear entirely.

Pixelate downsamples the region to a small grid, then draws it back up with smoothing switched off, which is what produces the familiar blocky mosaic look. It destroys dramatically more detail than a blur, there's no gradient to reconstruct, just flat colour blocks, but it still visually reads as 'there was a face here, now blocked out', which some people prefer over a plain black rectangle for readability of the photo as a whole.

Both options are one-way. Once a region is redacted and the photo is marked done, there's no slider to bring any of that detail back, which is the actual guarantee a redaction tool needs to make.

How to redact a photo, step by step

Open the redact and blur tool and drop your photos in, one or several. Photos go through one at a time rather than all at once, since redacting needs you to actually look at each photo and decide what to cover, there's no way to automate that safely without either missing something or covering something that didn't need it.

For the photo on screen, click and drag over whatever needs covering, a face, a plate, text on a document. Release, and the redaction is applied immediately, baked directly onto the photo. Draw as many boxes as the photo needs, then move to the next one.

When a photo is done, mark it done and move to the next. At the end, download everything as one ZIP, same as any other batch on this site.

Why this has to be manual, not automatic

Every other tool on this site runs the same operation on every file in a batch automatically, resize them all the same way, compress them all to the same target. Redaction can't work like that, because what needs covering is different in every single photo, a face here, a plate there, an address in the corner of a different shot entirely, and there's no safe way to guess that without a real risk of missing something that mattered or blacking out something that didn't.

So this tool keeps the batch queue, you can still load a whole folder at once, but each photo pauses for you to actually look at it and decide what to cover, rather than running a blind pass across everything. It's slower per photo than an automatic tool, and that's deliberate: the one thing a redaction tool cannot afford to get wrong quietly is missing the thing you needed hidden.

Multiple regions and getting them right

A single photo can need more than one box, a face and a plate in the same shot, or several faces in a group photo. Draw each one separately, they all apply to the same working copy of the photo, so by the time you mark it done every region you've drawn is baked in together.

Take a moment before marking a photo done to check the edges of each box. A box drawn slightly too tight can leave a sliver of the original visible along one side, which defeats the purpose of covering it in the first place. It costs nothing to draw a box a little larger than the strict minimum, better to cover a bit of surrounding background than to leave a hairline of the thing you were redacting.

Undoing a mistake

If you draw a box in the wrong spot before finishing that photo, undo removes the most recent region and redraws the photo from its original state plus every region before that one, rather than trying to un-redact a specific box out of order. Draw it again correctly and continue.

Once a photo is marked done, that's final for that photo, there's no undo after the fact, because the whole point of redaction is that the original detail is genuinely gone by then, not held onto in case you change your mind. If you're not confident about a box, check it before moving on rather than after.

Does the box on screen actually cover the full-resolution photo

Yes, and this is worth trusting rather than eyeballing. A large photo is very likely displayed smaller on your screen than its actual pixel dimensions, a four-thousand-pixel photo doesn't render at four thousand screen pixels. Every box you draw gets scaled up to the photo's real resolution before it's applied, and the redaction itself is baked onto that full-resolution version, not a shrunk preview.

So a box that looks like it covers a face on your screen covers that same face in the exported file at full size. This matters because getting it backwards, redacting a small preview and calling it done, is the one failure mode that would actually put someone's information back in the file you thought you'd cleaned.

Common reasons people redact a photo

Selling something online and photographing it where it sits, a car with the plate visible, a package with the shipping label still on it, a room with mail on the counter. Sharing a photo of kids where you'd rather their faces weren't identifiable to strangers scrolling past. A photo taken in public that happens to catch a stranger's face clearly enough to be recognisable.

Screenshots are a common one too, a chat, a form, a spreadsheet, where an account number, an address, or someone else's name appears somewhere in the frame and needs to disappear before the screenshot goes anywhere else.

Redacting a document or screenshot vs a photo

The tool doesn't distinguish between a photo of a physical scene and a screenshot, both are just an image with pixels in it, and the same fill or pixelate operation works identically on either. Drop a screenshot in the same way you'd drop a photo, draw a box over whatever needs covering, and it's handled the same way.

One thing worth knowing for documents specifically: pixelating text is less reliably unreadable than pixelating a face, because text has hard, high-contrast edges that can survive a pixelation pass better than the soft tones of a face do. For anything with actual text you need genuinely gone, an account number, a signature, a name, solid fill is the safer of the two options.

How this differs from just cropping the photo

Cropping removes a region by cutting it off the edge of the photo entirely, which only works when the thing you need gone happens to be at the border of the frame. A face in the middle of a group photo, or a plate on a car that's the actual subject of the shot, can't be cropped away without cutting out everything around it too.

Redaction covers a region in place instead, so the rest of the photo, its framing, its composition, everyone else in it, stays exactly as it was. It's the right tool whenever the sensitive part is somewhere inside the frame rather than conveniently sitting at an edge you could just trim off.

Redacting an already-redacted photo

If you run a photo you've already redacted back through the tool, there's nothing left to find in the areas you covered, that's the actual guarantee solid fill and pixelation make, so there's no partial information sitting there for a second pass to accidentally reveal. You'd only ever do this to cover something additional you missed the first time, not to strengthen a redaction that's already final.

That's a meaningfully different guarantee to a blur, which can sometimes be sharpened or deblurred further with more aggressive processing precisely because a blur never fully throws the original detail away in the first place. Fill and pixelation don't have that problem because there's no residual signal left to work with.

Why nothing gets uploaded to do this

The photos being redacted here are, almost by definition, the ones with something sensitive in them, a face, an address, a document. Sending that photo to a server to have the sensitive part covered is a strange trade to make, you'd be exposing the exact thing you're trying to protect to get it protected.

This tool runs entirely in your browser. The photo you drop in never leaves your device at any point, before, during, or after redaction. You can check that directly: load the page, disconnect from the internet, and redact a photo anyway. It still works, because there was never a server it needed to reach.

What this tool isn't for

This is a single-image, manual tool, not an automatic face-detection redactor that scans a batch and blacks out every face it finds without you looking. That kind of automation exists elsewhere, and it trades accuracy for speed, it will miss faces at odd angles or in poor light, and it has no way to know a plate or an address needs covering at all, since it isn't looking for either.

It's also not built for video. Redacting a moving face or plate needs the same box tracked and reapplied across every frame, a meaningfully different and heavier job than a single static image, and not something a browser tab is the right tool for.

Quick questions

Tools

• Redact & blur, hide a face, plate or address before you share a photo