Automatically Blur Faces in a Video
Before you post a video, automatically find and hide the faces of bystanders, kids, or anyone who didn't agree to be shown, using mosaic or blur. The video never leaves your device - everything runs inside this browser, nothing is uploaded to a server. This tool exists to help protect people's right to privacy in footage you share.
Unlike most "video face blur" tools online, this one does not require installing an app, does not use ffmpeg.wasm, and does not send a single frame to any server. Detection and rendering both happen with on-device machine learning and the browser's own canvas and recording APIs, so the whole process stays private and works offline once the page and its small AI model are cached.
Loading the tool...
How to use
- Drop in the video you want to edit, or click to choose a file. (MP4, MOV, WEBM · up to 3 minutes)
- Pick an effect (mosaic or blur), the strength, and how much margin to pad around each detected face box.
- Press "Find faces (preview)" to scan the video for faces. The first time this loads a small on-device AI model.
- Check the detected face boxes. Click a box to exclude a face you don't want blurred, and drag on the preview to manually add a box for any face the detector missed.
- Press "Apply mosaic & save". Rendering happens roughly in real time relative to the video length, so keep the tab open until it finishes.
- Play back the result to confirm every face is properly covered, then download it.
Why automatic face blur matters
Sharing videos from a trip, a street performance, a classroom, or a family gathering almost always means other people end up in frame who never agreed to appear online. Manually scrubbing through footage frame by frame to paint over faces in a video editor is slow and error-prone, especially once people start moving around. This tool automates the tedious part - tracking a face across dozens or hundreds of frames - while still letting you review and correct the result before anything is final.
Mosaic or blur, and how strength and padding work
Mosaic pixelates each detected face into large, visible blocks, which reads clearly as "this has been intentionally covered" even at a glance. Blur instead smooths the face into a soft, less distinct shape, which can look a little more natural in context. Strength controls how coarse the mosaic blocks are, or how strong the blur radius is - higher values protect privacy more firmly but look more obviously edited. Padding expands each face box outward by a percentage before the effect is applied, which helps make sure hair, ears, or a slightly-off detection box doesn't leave a sliver of face visible at the edge.
FAQ
Can a blurred video be reversed back to the original?
No. The original pixels under each blurred or pixelated area are discarded in the saved video. That said, very light pixelation or blur can sometimes let a shape be guessed, so we recommend a higher strength when privacy really matters.
Does it find every face automatically?
It uses an on-device face detector to find front-facing faces and mark them with boxes. Side profiles, small faces, and faces partly covered by a mask or hand can be missed. Always check the boxes in the preview, and add any missed face by hand before processing.
Is there a length or resolution limit?
Videos up to 3 minutes are supported. Anything larger than 1080p is automatically shrunk to 1080p while processing. For longer videos, trim them first with another tool.
Will the result be MP4 or WebM?
Recent Chrome, Edge and Safari versions usually save as MP4. Depending on the browser it may only be able to save WebM, which can fail to play on some older phones or inside messaging apps - if that happens, try again in an up-to-date desktop browser to get MP4.
Does the original audio stay in the result?
Yes, if you keep the "Keep sound" option checked, the original audio track is carried over into the result video. Uncheck it to save a silent video instead.
Is my video uploaded to a server?
No. Face detection and the blurring itself both run entirely inside this browser, on your own device. The video file is never sent anywhere.
Open source used
Face detection runs on-device using Google's MediaPipe Face Detector (Apache-2.0 license). The model file and the WASM runtime are both served directly from Dagochim's own servers, and no video frame or detection result is ever transmitted anywhere else.
⚠️ Face detection can miss side profiles, small faces, and faces that are partly hidden. Please watch the preview and the finished result carefully before sharing the video. This tool is meant to help protect people's privacy, not to guarantee a perfect result every time.