Pause a clip on one frame and a soft wash over a cheek can look fine. Play the same clip and the brain stitches dozens of slightly different angles into a person you recognise. That gap between a paused preview and motion is why face cover on video is a planning problem, not a one-tap stamp.
BlurThis keeps the work in your browser tab: timed boxes, Blur In and Blur Out, Blur or Pixelate at high strength, and an export encoded on your device. There is no automatic face tracking and no opaque black bar on video. The craft is choosing an oversized box when the head only sways, chaining short segments when someone walks the frame, and cutting the high-risk seconds when paint is the wrong answer.
Open larger image in a new tabHow to blur out faces in a video online free
Plan for a moving head. One oversized box for a sway; chained short segments for a walk. Raise Blur or Pixelate until the face is an even wash or large tiles at pause. Cut high-risk seconds. Limits: 50 MB in, 90 seconds out, longest edge 720 pixels.
In BlurThis
These steps are for a moving face on a short clip. Open the video blur tool, then:
- Choose a clip under 50 MB. Trim to the seconds you will share (export caps at 90 seconds; longest edge scales to 720 pixels).
- Scrub to the first frame the face is visible. Press Blur In. Drag a box with padding past hairline, ears, chin, and glasses.
- Scrub to the last frame that still needs cover. Press Blur Out. If the head drifts out, add another short segment where it landed and overlap the handoff by a few frames.
- Raise Blur or Pixelate until earrings, glasses frames, and hair edges stop reading at pause. For a close-up you cannot safely show, cut those seconds from the trim instead.
- Play the export and watch the face, not the box outline. Press Download and post that file.
Fixed boxes on a moving head
On video, a cover is a segment with a start, an end, a shape, and a strength. The box stays where you put it. It does not follow the skull. If you expect auto-tracking, you will leave gaps the moment someone leans, stands, or crosses the frame. If you plan for fixed boxes, a thirty-second clip finishes in a few careful passes and the rest of the picture stays sharp.
Video covers are Blur or Pixelate only. Letters, jersey numbers, name badges, and plates in the same shot need a heavy cover or a trim cut. Soft blur on digits is the wrong tool even when a face blur is acceptable. For general timed regions that are not faces (URL bars, toasts, plates), use how to blur part of a video online; this page stays on the moving-person problem.
Open larger image in a new tabOversized box versus a chain of segments
Movement decides the tactic. Pick one strategy on purpose instead of fighting a narrow rectangle for the whole clip.
One oversized box. When the person drifts inside a small area (talking-head sway, a child at a table, a seated interview), make a single box large enough to cover everywhere the face goes for that window. You will blur more background than strictly needed. You will not leave a cheek peeking out at the edge. An oval can match head shape and waste less of the picture than a tall rectangle.
Several short segments. When someone walks across the frame, chain covers: place one, set Blur In and Blur Out over a second or two, then place the next where the head has moved. Overlap by a few frames so there is never a gap at the handoff. Segment boundaries are where leaks happen. A single uncovered frame at the join is enough to identify someone, because anyone can pause.
Do not cover two independently moving people with one narrow box. Give each face its own segment. When paths cross, a wide box across both for the crossing, then separate boxes again once they part, is usually the cleanest fix. Step through that crossing in the export frame by frame.
Strength under motion
Push strength higher than you would on a still. A viewer gets many frames of the same face, and recognition averages them. A blur that looks adequate on one paused frame can still read as a person once the clip plays. Keep raising until earrings, glasses frames, and hair edges stop reading at full zoom, or enlarge the box.
Soft blur looks natural on a casual social post. Pixelate reads as deliberate censorship. They are different looks, not a ranked strength scale. Either works for a face at high strength; neither works at low strength. Face cover on video is still not identity-proof. Distinctive jackets, tattoos, and spoken names survive a washed face.
When cutting beats covering
Some moments should not be painted. A close-up of a child who is not yours, a customer next to a readable badge, or a few seconds where someone turns into the lens and fills half the frame are often cleaner as a trim than as a heroic blur. Cutting removes the pixels. A weak cover leaves a puzzle for anyone who pauses.
Use cover when the surrounding action must stay. Use cut when the only reason those seconds exist is the person you cannot safely show. Mixing both (trim the worst beat, then box the rest) is normal editorial judgment. Remember the hard limits while you plan: files over 50 MB are rejected, export stops at 90 seconds, and the longest edge scales to 720 pixels. MP4, WebM, and MOV names are accepted; an extension does not guarantee every codec. Chrome and Edge on a desktop remain the most reliable path for MP4 with audio.
Check the download, listen as well
Scrub the downloaded file at normal speed and watch the target. Then step through the first and last frames of each segment. Do not screen-record the editor preview; that shortcut picks up notifications and adds a second compression pass.
Picture cover does not remove a name spoken aloud, a burned-in caption, or a distinctive voice. Listen through. If a name is spoken, cut the audio in another app or trim the moment. Audio is usually kept on export; some browser and codec combinations drop it and give you a video-only file with a notice. Closing the tab drops the working copy from this page. That is not a forensic wipe of disk, backups, or chat history. Editing stays in the tab; nothing is sent to BlurThis servers for processing.
Situations that show up every week
- Kids and classrooms. Cover every child who is not yours, then look for names on bags, jerseys, and the board. A face cover is unfinished if the jersey still prints the name.
- Bystanders on a street. A wide, heavy box as they pass is faster than a new segment for each person; most are on screen for under two seconds.
- Customers and colleagues in a demo. Cover the face, then a second timed box on the nameplate or badge. Screen-recording leaks (URL bars, toasts) belong on the part-of-video guide linked above.
- Yourself. A face blur plus a distinctive jacket, tattoo, or front door is not anonymous. Cover the second identifier or crop it out of the trim.
What this tool does not do
It does not track a face automatically. It does not paint an opaque black bar on video. It is not a full NLE. It does not render in the cloud. Those are the trade-offs of keeping the encode on your device. For most clips people actually share (a face for a few seconds, a bystander for two), they are the right trade-offs.
For a still-photo face job with optional Find faces, see blur a face in a photo online free. Wider map: how to blur hub. Still face tool: blur a face.
FAQ
Is the video uploaded? No. The clip is decoded, covered, and re-encoded inside the browser tab on your device, then saved to downloads. Delay while it encodes is not proof of an upload. A Network panel check should include images, media, and other requests, not only Fetch/XHR. Offline use after load shows local capability; it does not prove an online session never loads other page assets. Stay on the video editor.
Will it keep the audio? Usually. Some browser and codec combinations cannot mux audio; you then get a video-only file and a notice. Prefer Chrome or Edge on a desktop when you need sound.
Can I blur a face in a photo the same way? Yes, and it is easier: no timeline, and Find faces can suggest boxes. Use the still guide linked above rather than forcing a video workflow onto a PNG.
