Filming on a street, in a cafe, at a playground, or at an event almost always means other people end up in the shot - a stranger walking by, someone else's kid, a person sitting at the next table who never agreed to appear in your video. Posting that footage as-is on YouTube or social media can violate their right to privacy, and can lead to takedown requests or worse. Opening a video editor and manually painting over a moving face frame by frame is, frankly, a miserable way to spend an afternoon.
Why blurring faces in video is so much harder than in a photo
Covering a face in a single photo and covering a face in a video are completely different problems. A photo only needs one pass - draw a box once and you're done. A video is a different story because people keep moving, so the face's position changes on every single frame. A 10-second clip at 30 frames per second already has over 300 frames to track through, and following a moving face by hand across all of them is slow even for a professional editor. That's usually why people either pay for software with a "motion tracking" feature, or just give up and reshoot the scene so no one else's face is visible.
What Dagochim's automatic video face blur tool actually does
Dagochim's Automatic Face Blur for Video tool runs Google's open-source face detection model (MediaPipe Face Detector) directly inside your browser to find faces in the video and track their movement, applying mosaic or blur to follow them through the clip. There is no server upload step at all - family footage or anything sensitive is processed entirely on your own computer or phone and never transmitted anywhere else.
Step-by-step guide
- Open the Automatic Face Blur for Video tool.
- Drop in an MP4, MOV, or WEBM video. (Up to 3 minutes; anything larger than 1080p is shrunk automatically.)
- Choose an effect - Mosaic or Blur - and set the strength and the margin padded around each detected face.
- Press "Find faces (preview)". The first time this runs, it downloads a small on-device AI model (about 0.2 MB).
- Review the face boxes shown on screen. Click a box to exclude a face you don't want covered (such as your own), and drag over 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's length, so keep the tab open until it's done.
- Play back the finished video to confirm every face is properly covered before downloading it.
Rendering runs in real time relative to the video's length - a 3-minute clip can take around 3 minutes to process. If you switch to another tab, it pauses automatically and picks back up once you return.
Mosaic vs. blur: which should you pick
Mosaic turns each detected face into large, visible blocks, which makes it immediately obvious to any viewer that the footage was deliberately censored. That's a good fit for news footage, submissions, or anything where it needs to be unmistakable that a face was hidden on purpose. Blur instead softens the face into an indistinct shape, which tends to feel less jarring and fits more naturally into a vlog or everyday clip where you don't want to disrupt the viewing experience too much. Neither option is "more private" than the other at the same strength - it's really a stylistic choice about how obvious you want the edit to look.
How strength and padding actually work
Turning up the strength makes mosaic blocks larger or the blur radius stronger, which makes the original face harder to make out. If you need to make sure a stranger's identity is properly hidden, lean toward a higher strength rather than a lower one. Face box margin pads each detected box outward by a percentage before the effect gets applied, which helps make sure hair, an ear, or a slightly-off detection box doesn't leave a sliver of untouched face peeking out at the edges. A bit of extra padding is cheap insurance against a mostly-good detection box that's just a few pixels too tight.
Cases the face detector can miss
The on-device detector is good at finding clear, front-facing faces, but it can miss faces in several common situations:
- Side profiles or faces turned away from the camera - detection accuracy drops sharply once a face isn't roughly facing forward.
- Faces that are very small in the frame - someone far from the camera takes up very few pixels, which makes them easy to miss.
- Faces covered by a mask, sunglasses, or a hand - when key facial features are hidden, detection gets much harder.
- Very dark scenes or strong backlighting - heavy contrast can obscure the facial outline the model relies on.
That's exactly why this tool always shows a preview step where you confirm the detected boxes and can manually add any that were missed, rather than silently trusting the detector and rendering straight through. Don't rely on the automatic pass alone - make a habit of watching the finished result all the way through before you share it, specifically checking for any face that slipped past detection.
Will the output be MP4 or WebM?
Most up-to-date versions of Chrome, Edge, and Safari will save the result as MP4 (H.264), which plays back on nearly any device. Depending on the browser, though, it may only be able to save WebM instead, and WebM can fail to play on some older phones or inside certain messaging apps. If that happens, try the same video again in an up-to-date desktop browser to get an MP4 file instead.
What happens to the audio?
Keeping the "Keep sound" option checked carries the original audio track straight through into the result video - useful when you only need to hide faces but want to keep the sound exactly as it was. If the audio itself is sensitive and needs to go too, simply uncheck that option to save a silent video instead.
Do you actually need to do this? A note on privacy and the right of publicity
Many jurisdictions recognize some form of a person's right not to be filmed or have their likeness published without consent, often called the "right of publicity" or similar privacy rights, especially once someone is identifiable rather than an anonymous blur in the background. A bystander caught clearly on camera on a public street can still have grounds for a complaint if their face is recognizable and the footage is published without consent. This matters even more for children's faces, or for other customers caught in the background inside a private space such as a cafe or restaurant - when in doubt, it's safer to blur.
Alternative: motion tracking in a video editor
Professional tools such as Premiere Pro, DaVinci Resolve, or CapCut include a "tracking" feature that follows a face across frames and applies a blur effect to it. If you need finer control, or you're already editing the video professionally for other reasons, those tools can offer more precision. But they come with a learning curve, an install, and often a long render time. For a quick one-off job on one or two clips, doing it directly in the browser with Dagochim is considerably simpler and faster to get to a finished file.
Common problems and fixes
- An iPhone video (HEVC .mov) won't open - if a browser can't play a video, it can't process it either. Try Safari, or a recent version of Chrome/Edge with HEVC support.
- No faces were detected at all - this can happen with dark lighting or faces that are very small in frame. Drag over the preview to add a box by hand instead.
- The tab was closed partway through processing - the in-progress render is lost when that happens, and you'll need to start over from the beginning.
Is this actually private?
Yes. Dagochim's video face blur tool never uploads your video to a server - face detection and the blurring itself both happen entirely inside this browser, on your own device. Even sensitive family footage never leaves your computer or phone.
A quick checklist before you share the result
Before publishing the finished video anywhere, it's worth running through a short checklist: watch the entire clip at normal speed once, specifically looking for faces near the edges of the frame or in the background; check any scene where people walk quickly across frame, since fast motion is where tracking is most likely to lag a box slightly behind; and double-check any side-profile shots specifically, since those are the single most common type of face the detector misses. Taking two extra minutes to do this review is far cheaper than having to re-process and re-upload a video after someone points out a missed face.
Frequently asked questions
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 use 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.