Privacy Blur
Blur faces in a photo before you share it. A small open-source detector finds faces in your browser — the photo is never uploaded — and you can hide them with a blur or a pixelation. Drag a box on the picture to cover anything it missed.
Private by design: Faces are detected by a small model running in your browser; your photo is never uploaded. The model is downloaded once from Hugging Face and cached on your device. See how in-browser processing works.
How to blur faces in a photo
Drop a JPG, PNG, or WebP onto the upload area, or click to browse.
Press Find faces. The detector — just over 1 MB — downloads once and runs in a fraction of a second.
Check every face is covered. Drag on the original to add a box the detector missed; click any box's label to remove it. Choose blur or pixelate, set the strength, and download the PNG.
Why blur faces at all
Sharing a photo shares everyone in it. A picture of your team, a crowd at an event, a classroom, a street scene, or a screenshot of a video call can include people who never agreed to appear online — colleagues, children, bystanders. Blurring their faces lets you share the moment without sharing their identity. It's also the simplest way to post a photo of your own child that friends can enjoy but strangers can't search.
Screenshots are a common case people forget: a group video call, a social feed, or a news article in the background of a photo can carry a dozen recognisable faces you didn't intend to publish.
How the detector works, and what it misses
The tool uses UltraFace, a small open-source face detector released under the MIT licence. It runs entirely in your browser — about 35 milliseconds per photo on an ordinary laptop, with no graphics chip needed — and finds faces looking roughly toward the camera.
No detector finds every face, and this one is no exception. It is most likely to miss faces that are very small in the frame, turned mostly away, partly covered by hands, hair, or objects, heavily shadowed, or in low-resolution or blurry photos. That is exactly why the tool lets you draw boxes by hand: always check the result yourself before sharing. The count it reports is what it found, not a guarantee of who is in the picture.
Blur versus pixelate
Blur softens the face into a smooth patch; pixelate turns it into large blocks. Both hide identity at medium and strong settings. Light settings leave a face recognisable to someone who already knows the person — treat them as a stylistic choice, not a privacy one. If the goal is that nobody can identify anyone, use medium or strong.
One honest caveat about all face hiding: it obscures the face, not the person. Clothing, setting, body shape, and who else is in the photo can still identify someone to people who know them. For photos where that matters, consider whether to share them at all.
Good to know
Everything is processed on your device. The detector downloads once from Hugging Face and is cached; the photo itself is never sent anywhere.
Detected boxes are widened slightly so hairlines and chins are covered. Manual boxes are exactly where you draw them.
Output is a PNG at the original size. To share a smaller file, run it through the Image Compressor afterwards.
To soften a whole background rather than specific faces, use Background Blur; to remove the background entirely, the Background Remover.
How to blur faces in a photo before sharing it covers when to do it, what the detector misses, and the limits of face-hiding.
Frequently asked questions
Is my photo uploaded to find the faces?
No. The detector runs in your browser and the photo never leaves your device. The only network traffic is the 1.2 MB model file, downloaded once from Hugging Face. We deliberately chose a detector that runs on our own on-device machinery rather than a library that reports usage back to its maker.
Does it find every face?
No, and no automatic detector does. It is most likely to miss faces that are small, turned away, partly covered, in shadow, or in low-resolution photos. Check the result and drag a box over anything it missed — that is what the manual boxes are for.
Can I hide something that isn't a face?
Yes. Drag a box over any area of the original — a name badge, a licence plate, a screen, a document — and it is blurred or pixelated the same way. The detector only finds faces; the boxes hide whatever you choose.
Is blurring enough to make someone unidentifiable?
At medium or strong strength the face itself is not recognisable. But a face is not the only identifier: clothing, location, and the other people present can still reveal who someone is to people who know them. Blurring is a strong step, not a guarantee.
Can a blurred or pixelated face be reversed?
Research has shown that light pixelation and light blur can sometimes be partly reversed. Use medium or strong for anything you genuinely want hidden — at those settings the original pixels are gone from the saved file, not merely obscured on screen.
Which model is this, and can I trust it?
UltraFace (Linzaer's Ultra-Light-Fast-Generic-Face-Detector), an open-source detector under the MIT licence with published code and weights. It was chosen over a better-known alternative because that alternative sends usage metrics to its maker with no way to turn them off — not appropriate for a privacy tool.
Guides for this tool
- How to Blur Faces in a Photo Before Sharing It — The photo is good and you want to post it. It also has your kid's classmates in it, or colleagues who didn't sign up to be on your feed, or a stranger who happened to be in frame. Blurring their faces lets you share the moment without publishing their identities. Here is how to do it properly — including the part where you check the tool's work.
Related tools
Privacy Blur is one of 4 AI tools — remove or blur backgrounds, hide faces, and upscale — models run on your device.