This is the MozFace team. Hiding a face in a photo takes a second, but the moment you switch to video it turns into real work. Faces move, and even a short clip is hundreds of images in a row. This article walks through how to blur faces in a video — keeping the mosaic on a moving face, covering things that are not faces, and exporting the finished clip.
Why video is harder than photos
The main reason is that your subject moves. In a photo a face sits in one place. In a video people walk and the camera shakes, so a mosaic that lines up perfectly at one moment is covering empty pavement a few seconds later — with the face fully exposed.
Most video editors handle this with keyframes: you mark the position at 0 seconds, mark it again at 3 seconds, and the app fills in the movement between them. Done carefully it is accurate, but as soon as someone changes direction, or several people move in different directions, the number of points you have to place keeps growing.
The second reason is that checking your work takes time. A missed spot in a photo is obvious at a glance. A video has to be played all the way through. Realising after you posted that a face was visible for a few seconds in the middle is the failure worth designing against.
Blurring a video in MozFace, step by step
MozFace automates the "follow the moving face" part. There are four steps.
Add your video
Open the app and pick a video from your device. Face detection starts as soon as it loads.
Faces are detected and tracked
Every face is detected and its position through the clip is tracked over time. Detected faces appear already blurred, all of them.
Reveal only who you want to show
Detected faces are listed as thumbnails. Tap yourself or anyone else you are happy to show, and that person alone is unblurred.
Preview, then export
Play it back to check the result, then save. The edited video is written to your gallery and the original file stays untouched.
Step 3 is the important one. Because everyone starts out hidden, the job becomes "choose who to show" rather than "find everyone to hide." Exposed faces happen when you forget to hide someone. If everyone is hidden from the start, there is nothing left to forget.
The mosaic follows the face
For each detected face, MozFace records how its position changes over the length of the clip. If someone walks from one side of the frame to the other, the mosaic travels with them. You do not place a single keyframe.
The same holds when several people are moving in different directions at once. Each face is tracked independently, so there is no going back to reposition them one at a time. The more people are in the clip, the bigger that difference gets.
Covering license plates and other non-faces
Faces are not the only thing worth hiding. License plates, house numbers, signage, information on a monitor — none of that registers as a face, so you place it by hand. MozFace gives you three shapes for video:
- Oval — for faces and rounded objects, where a soft outline looks natural
- Rounded rectangle — for license plates, signs, and other boxy subjects
- Horizontal band — spans the frame, useful for a line of text or a lower-third caption
Each region can also be given a time range. If a sign is only visible from 0:12 to 0:18, the cover applies to just those seconds. There is no need to obscure the whole clip because of one moment.
Mosaic styles and strength
Video offers two styles: blur and pixelation. Blur dissolves edges smoothly and blends into the footage without breaking the mood. Pixelation paints over the area in square blocks — the familiar look that reads clearly as "this was hidden on purpose."
If you are unsure, decide by what you are hiding. For scenery and everyday footage you want to keep looking natural, use blur. For anything that must not be readable — plate numbers, text, documents — turn pixelation up. For photos, the app adds emoji stamps as a third style.
Your video never leaves the device
Most web-based video blurring services upload your file to a server to process it. That means footage you are editing precisely because it contains faces or personal details gets sent out first, unedited.
MozFace runs detection, mosaic rendering, and the final export entirely on your device. Your video is never sent to an external server, and the app works the same in airplane mode or with no signal. There is also no waiting for a large file to upload.
Check this before you post
Play the whole clip once before exporting. Pay attention to the moment someone crosses the frame, sharp camera movements, and a face turning back toward the camera. Remember that faces are not the only exposure: a house number in the background, a document on a table, a notification on a screen. Whether the finished video is appropriate to publish is your own call to make.
MozFace
Select a video and every face is detected and blurred automatically, tracked as they move. Tap to reveal only who you want to show. Everything runs on your device, fully offline.
Learn more about MozFaceFrequently asked questions
Is video face blurring free?
MozFace is free to download and use. Detecting faces in a video, tracking them, applying the mosaic, and exporting all happen inside the app.
Does the mosaic follow a face as it moves?
Yes. Each detected face has its position tracked across the clip, so the mosaic moves with the subject. You do not place keyframes by hand.
What if there are several people in the video?
Every detected face is tracked independently. All of them start out blurred, so you only tap the ones you want to reveal — which is why crowded clips rarely end up with someone missed.
Can I cover license plates and other non-faces?
Yes. Place an oval, a rounded rectangle, or a horizontal band by hand. You can also set which seconds it covers, so it can apply to just one part of the clip.
Is my video uploaded anywhere?
No. Detection through export runs entirely on your device. Your video is never sent to an external server, and the app works offline.
Is the original video kept?
The edited video is saved as a separate file, so the original stays exactly as it was.