Blur Faces in Video
Blur Faces in Video finds and tracks every face in a video with an AI model in your browser, blurs or pixelates them and exports an MP4, so the video is never uploaded.
At a glance
- Accepts
- MP4, MOV, WebM and MKV videos your browser can decode (HEVC needs Safari or a recent Chrome/Edge)
- Outputs
- An H.264 MP4 with faces blurred, pixelated or blacked out; chosen faces can stay visible; sound copied or re-encoded
- Where it runs
- Entirely in your browser, on your device
- What leaves your device
- Nothing - the file never leaves your device
- Limits
- Can miss side profiles, turned, covered or very small faces - check before sharing; first use downloads a 4.6 MB detector; up to 30 min on computers, 5 min and 720p on phones; needs WebCodecs
- Account
- Not required - there is no sign-up
- Price
- Free, with no watermark and no usage cap
- Made by
- Visrut
- Last updated
- 12 October 2026
Drop in a video and every face in it is found, followed from frame to frame and blurred, pixelated or blacked out, then saved as an MP4. Before anything is encoded you see each face it found and a few preview frames, and you can click any face to keep it visible. It all runs in your browser: the face detector (Google's MediaPipe BlazeFace) and the video encoder work on your device, and the video is never uploaded.
How to use it
- Drop in a video (MP4, MOV, WebM or MKV), or click to choose one.
- Press Find faces. Standard checks 10 times a second; Thorough checks every frame.
- Look over the faces and preview frames. Click any face you want to leave visible, and pick Blur, Pixelate or Solid block.
- Press Blur faces and export, then watch the result before you share it.
How it keeps the blur on the face
The detector looks at the frame as a whole and in overlapping half-size tiles, so it finds smaller faces than a single look at the whole frame would. Each face is then followed as a track: between two checks its box moves in a straight line, and if the detector loses it for a moment (a blink, a hand in front, a turned head) and finds it again nearby within three seconds, the gap stays covered. The box is also held a moment before a face is first found and after it is last seen, and grown by the Cover area setting, so the blur doesn't flicker off at the edges of a shot.
Honest limits
- It can miss faces: side profiles, faces turned away, faces partly covered, motion-blurred faces and very small faces in a crowd are the usual ones. A face that is never detected is never blurred, so check the result before you share it.
- Only faces are covered. Name tags, number plates, tattoos, screens and voices are left as they are.
- Every frame is decoded twice and re-encoded, so it often takes longer than the video plays. Phones handle up to 5 minutes at up to 720p; computers up to 30 minutes.
- For still photos, use Blur Faces, which also lets you draw boxes by hand. To make the result smaller for WhatsApp or Discord, run it through Compress Video; to cut out a part where a face was missed, use Trim Video.
Frequently asked questions
Is my video uploaded?
No. The video is decoded, scanned for faces, blurred and re-encoded inside your browser tab. Face detection uses Google's MediaPipe BlazeFace model running on your device. The first time, your browser downloads the detector (about 4.6 MB: the MediaPipe runtime from the jsDelivr CDN, or unpkg as a fallback, and the model file from Google's model storage) and caches it. The video itself never leaves the page.
Will it catch every face?
Not always, so watch the result before you share it. The detector finds faces looking roughly towards the camera well. Side profiles, faces turned away, faces partly hidden by hands, masks or hair, motion-blurred faces in fast pans and very small faces in a crowd are the ones it misses most. When a face is lost for a moment and found again in the same place, the gap is covered too, but a face that is never detected is never blurred.
It missed a face. What can I do?
Try Thorough mode: it checks every frame and looks closer for small faces. Increase Cover area so the blur reaches further around each face. If a face is still missed, trim that part out with Trim Video, or cut the video into shorter pieces where the face is larger. There is no way yet to draw a blur box by hand in a video.
Can I keep one person's face visible?
Yes. After the scan, every face it found is listed with a thumbnail, and the preview frames show a box on each one. Click a face or its box to keep it unblurred. The choice applies to that face for as long as it is tracked; if the same person leaves and comes back later, they may show up as a second entry to click.
Can the blur be undone?
Not from the exported video. Blur and Pixelate first shrink each face to between 4 and 16 pixels across and then scale it back up, so the detail is thrown away rather than softened. Solid block replaces the face with black.
What happens to the sound?
AAC sound, the usual kind in MP4 and iPhone videos, is copied over untouched. Other formats (Opus in WebM, for example) are re-encoded to AAC, or to Opus in Firefox, which has no AAC encoder. You can also remove the sound. Voices are not changed, so someone can still be recognised by their voice.
Which formats and browsers work?
MP4, MOV (including iPhone videos), WebM and MKV that your browser can decode. You get an H.264 MP4. It needs WebCodecs: Chrome, Edge and Safari 16.4+ on computers and phones, and Firefox 130+ on computers. HEVC (H.265) files need Safari, or Chrome or Edge on a computer with HEVC support. HDR iPhone video is converted to standard dynamic range so the colours stay right.
How long can the video be?
Up to 30 minutes on a computer and 5 minutes on a phone, where the output is also capped at 720p. Scanning and encoding run in a background worker, so the page stays usable and it keeps going if you switch tabs on a computer. On a phone, keep the tab in front.