Blur faces & licence plates in a video
Blur, pixelate or black-bar every detected face — and draw your own censor zones over licence plates, screens or anything else the detector can’t know about. Everything runs on your own device, so the video never leaves your browser.
How the face censor works
Every frame is scanned by an on-device face detector (Google's open-source MediaPipe BlazeFace model) while the video plays through once and is re-recorded. Detected faces are tracked from frame to frame with smoothing, so the censor region follows movement without flickering and stays in place for a moment when a face is briefly missed. Everything — detection, drawing and recording — happens in your browser; no frame is ever uploaded. Your own censor zones ride the same pass: draw a rectangle or oval on the preview — over a licence plate, a screen, a name tag — and it is censored in the same style, for the whole video or only between the times you set.
Is my video uploaded anywhere?
No — the face detector itself runs inside your browser, so even the frames being scanned never leave your device. That matters for exactly the kind of sensitive footage this tool exists for.
Can it miss a face?
Yes — faces that are very small, strongly turned away, badly lit or partly covered can be missed for a few frames. The tracker bridges short gaps, but always play the result back and check it before sharing.
Why does the first run download a few megabytes?
The detector engine and face model (about 11 MB) are loaded from our own server on first use and cached by your browser afterwards — no third-party servers are involved.
Can it censor licence plates and other things that aren’t faces?
Yes — draw a censor zone over it on the preview and it is blurred, pixelated or blacked out for the whole video, or only between the times you set. A zone stays where you draw it, so it suits anything that doesn’t move much; for a car that drives through the frame, give the zone its own short time window and make it generous.
Blurring faces properly
There are good reasons to blur a face that have nothing to do with hiding something. Footage shot in a public place almost always catches bystanders who never agreed to appear in it, and in Europe a recognisable face counts as personal data. Blurring the people who are not the subject is the simple way to publish a clip without chasing twenty strangers for permission. It matters in the other direction too — a witness, a patient, a child, a colleague who agreed to speak but not to be recognised.
Which style should you pick?
Blur is the softest and reads as an editorial choice rather than a redaction, which suits a busy street scene. Pixelate is unmistakable: the viewer sees that something was deliberately covered, and that is what you want when the point is that someone was protected. The black bar over the eyes covers less of the frame and keeps expression visible, which is why documentaries use it — but it also hides less, because hair, build and surroundings still identify a person to anyone who knows them. Where anonymity genuinely matters, use blur or pixelate at full strength over the whole head.
Getting the best detection
The detector wants a face that fills a reasonable part of the frame, roughly turned towards the camera and reasonably lit. Small faces in the background, strong profiles, heavy shadow, sunglasses and masks are where it can drop out for a few frames. The tracker bridges short gaps deliberately, but the rule stands: play the result back at full size before you share it, and watch the moments where someone turns or walks past the camera. If one person is missed again and again, a version of the clip cropped tighter on them will usually work.
