How long does processing take?

It depends on length, resolution and frame rate. Detection runs the AI model every few frames, and rendering decodes, blurs and re-encodes every frame: a 720p, 30 fps video takes roughly 1–3 seconds per second of footage on a typical computer. Trim the clip to the part you need first and cap the output at 720p to save time.

Some faces weren’t blurred. What can I do?

Raise the sensitivity, shorten the detection interval to every 2–3 frames, or enable small-face detection and detect again. Profiles, faces looking down, backlit faces and masks are harder to detect. For anything still missed, click Add a region by hand, drag a box over it in the preview and set its start and end times.

How do I blur only bystanders and keep myself visible?

After detection, scrub to a moment where that person appears and click their face box in the preview (or their track chip below it). The box turns into a green dashed outline, meaning it will be kept. If the tracker lost the person midway, they may appear as two tracks — set both to Keep.

Is my video uploaded anywhere?

No. The face detection model, frame decoding and FFmpeg encoding all run inside your browser, so the video never leaves your device — which matters for privacy-sensitive footage. The first run downloads the roughly 230 KB face model and the FFmpeg runtime, which are then cached by the browser.

🫥

🫥Blur Faces in Video Online - Auto Face Tracking & Pixelate

Need to hide bystanders or anyone who didn’t agree to appear before you post street footage, a meeting, a classroom or an interview? Upload a video and the tool runs the YuNet face detector every few frames, links detections of the same person into tracks with IoU tracking, interpolates the frames in between and keeps covering a moment before a face appears and after it leaves, so nothing flashes through. Scrub the preview to see the result at any moment: every face track is outlined, and clicking a face switches it to Keep — perfect for blurring passers-by while the presenter stays visible. Missed something like a profile, the back of a head or a licence plate? Draw a region by hand and set when it starts and ends. Choose blur, pixelate, black box or an emoji sticker, ellipse or rectangle, and adjust coverage and strength; detection sensitivity, interval, tiled detection for small distant faces and filtering of one-off false hits are all adjustable. Finally every frame is rendered and encoded to H.264 MP4 with FFmpeg, keeping the original audio, with optional resolution and frame-rate caps to save time. It all runs in your browser, and your video is never uploaded.

Free · No sign-up · Runs in your browserLast updated: 2026-10-26

Loading tool...

How to use Blur Faces in Video

1Upload a video, set the detection interval and sensitivity, and click Detect faces
2Scrub the preview: click faces to toggle blur / keep, add manual regions if needed and choose an effect
3Pick the output resolution, click Render video and download the MP4

FAQ

How long does processing take?: It depends on length, resolution and frame rate. Detection runs the AI model every few frames, and rendering decodes, blurs and re-encodes every frame: a 720p, 30 fps video takes roughly 1–3 seconds per second of footage on a typical computer. Trim the clip to the part you need first and cap the output at 720p to save time.
Some faces weren’t blurred. What can I do?: Raise the sensitivity, shorten the detection interval to every 2–3 frames, or enable small-face detection and detect again. Profiles, faces looking down, backlit faces and masks are harder to detect. For anything still missed, click Add a region by hand, drag a box over it in the preview and set its start and end times.
How do I blur only bystanders and keep myself visible?: After detection, scrub to a moment where that person appears and click their face box in the preview (or their track chip below it). The box turns into a green dashed outline, meaning it will be kept. If the tracker lost the person midway, they may appear as two tracks — set both to Keep.
Is my video uploaded anywhere?: No. The face detection model, frame decoding and FFmpeg encoding all run inside your browser, so the video never leaves your device — which matters for privacy-sensitive footage. The first run downloads the roughly 230 KB face model and the FFmpeg runtime, which are then cached by the browser.