Sitr

Blur the faces in a video without uploading it.

Drop in an MP4, a MOV or a WebM. Every face is found and the eyes covered — through movement, turns, and the frames where a detector would normally lose it. It all happens on this device, so the file is never uploaded to anybody, including us.

Free, no limitsNo accountNothing uploaded

Covered on this device0 bytes uploaded
The cover sits inside the same movement as the face. It does not lag behind a turn, and it does not slip off in the frames where the face is half in shadow.

It is not a blur

A blur averages neighbouring pixels and keeps the face in the file, smeared. This replaces the pixels. Why that matters.

It holds through the turns

Detection alone drops out at steep angles and in shadow. Tracking fills the gaps, and where it cannot, the bar widens across the eyes rather than disappearing.

Sized to the eyes

The cover follows the eyes frame by frame, so it stays a bar across them through fast movement and close-ups instead of swelling over the whole face.

How it goes

  1. 01

    Drop the video in

    MP4, MOV or WebM, at whatever resolution your camera produced. Nothing is sent anywhere when you do this — the browser is simply opening a local file.

  2. 02

    Every face is found and covered

    On your own hardware, with the same video engine your camera app uses. Each face is tracked so the cover moves with it rather than being pinned to a spot on the screen.

  3. 03

    The result is checked again

    The written file is decoded and re-read frame by frame by a stricter pass. Anything it finds is covered again, wider, and anything it cannot clear is shown to you with the timestamps.

  4. 04

    Download it, or post it

    Save it, or send it straight to TikTok, Instagram, YouTube and Facebook with one caption. Even then it goes from your browser to the platform directly and never passes through us.

What it takes, and what it gives back

Formats in
MP4, MOV, WebM — anything your browser can decode.
Resolution out
The one you gave it. No downscale and no watermark.
Covers
Frost, bar or streak. All three destroy the pixels underneath; none of them is a blur.
Where it runs
Chrome, Edge, or Safari 16.4 and later — on a laptop or a phone. It works with the network off.
What we receive
Nothing. There is no server in the path, so there is nothing to leak, breach or hand over.
Cost
Free. No limit, no account, no card.

Why it covers rather than blurs

Blurring is a weighted average of neighbouring pixels, which means the face is still present in the file in a spread-out form. The transform is close to invertible, and there are published attacks that walk it backwards far enough to recognise somebody — and a second kind of attack that does not need to reverse anything at all, only to compare your blurred face against a short list of candidates.

Replacing the pixels outright leaves nothing to reverse and nothing to compare. Which is why this ships coarse covers and solid bars, and will not ship a blur, whatever the address of this page says. The long version, with the maths.

Questions

Is the video uploaded to a server?

No. The file is opened, decoded, covered and re-encoded inside the browser tab, using the same video hardware your camera app uses. There is no upload step anywhere in the process — you can disconnect from the network before you start and it finishes exactly the same, which is the easiest way to check the claim rather than take our word for it.

Does it work on a phone?

Yes, in a recent Chrome or Edge, or Safari 16.4 and later. A long clip takes longer on a phone than on a laptop because it is doing real work on real hardware, and the screen gives you an estimate before you commit to it rather than after.

What happens when someone turns away from the camera?

The cover stays on. A detector loses a face at steep angles, in shadow and during fast movement, so tracking fills the gaps between detections and extrapolates briefly past the end of one. Where detection has been lost for longer, it fails closed — the bar widens into a band across the eyes rather than disappearing.

How do I know it did not miss a frame?

It does not sample. Faces are looked for on every frame — every other frame at 60fps, with the one between filled from both sides — and tracked from one frame to the next, so a fast turn or a lean into the lens does not leave a gap. The finished video also plays on the same screen before you download or post it, so you can watch it through first.

What formats and resolutions does it take?

MP4, MOV and WebM in, at the resolution you gave it. There is no downscale to 720p and no watermark, and no limit on how many — it is free.

Can I do several videos at once?

Yes. Drop in a batch and they run one after another with a single set of settings, and anything questionable in any of them is flagged at the end for a look.