How to blur faces in a video when people keep moving

To blur a face in a video, place a mask over the face, track it through every frame where the person is visible, and inspect the export around entrances, exits, cuts, and occlusion. The difficult part is not drawing the first mask. It is keeping coverage accurate while the person moves.

Zight does not currently claim automatic face detection for post-recording video redaction. Use an approved editor with manual masking or face tracking, then treat the complete exported video as the item that must pass review.

Choose the face coverage your audience needs

A soft blur may be enough when you only need to reduce distraction. Strong privacy redaction usually requires a denser blur, pixelation, or an opaque cover that prevents recognition. Follow the policy that applies to the recording rather than choosing the least noticeable effect.

Include more than the central features of the face. Hairline, profile, tattoos, name badges, reflections, and distinctive clothing can still identify someone even when the eyes and mouth are obscured.

Track one face from entrance to exit

1. Move to the first frame where the face becomes visible and draw a mask with margin around the head.
2. Start motion tracking if the editor supports it. If not, add keyframes often enough to keep the mask centered as the person moves or changes size.
3. Pause at turns, fast movement, camera zoom, and partial occlusion. Correct the mask before continuing the track.
4. End the mask only after the face has fully left the frame. Check the final visible frame and the next frame.
5. Restart with a new track after a hard cut or when the person re-enters in a different position.

Handle several people without losing track ownership

• Use one mask and one track per person so corrections do not move every cover at once.
• Name or color-code tracks inside the editor if it supports labels.
• Check crossings where two faces overlap and a tracker may jump from one person to another.
• Inspect picture-in-picture windows, mirrors, thumbnails, video-call participant strips, and background monitors.
• Search later scenes for the same person instead of assuming one track covers the full recording.

Correct tracking around occlusion and fast movement

Automatic motion tracking can lose a face when the person turns away, walks behind an object, leaves the frame, or becomes small. Stop at the last accurate frame, correct the mask manually, and resume from a clean reference frame.

For motion blur or a whip pan, enlarge the mask rather than letting identifiable features escape between keyframes. The final treatment should follow the full visible area, not the tracker’s smallest bounding box.

Review the face-blur export frame by frame

• Check every entrance and exit at reduced playback speed.
• Pause on cuts, transitions, zooms, and moments when subjects cross.
• Look for faces in reflections, thumbnails, posters, and secondary windows.
• Confirm the mask remains large enough at the nearest camera position.
• Review audio, captions, lower thirds, and filenames for names that identify the person.
• Test the final file on a second screen and at the resolution viewers will receive.

What Zight can and cannot automate today

Zight’s account-enabled post-recording workflow detects supported sensitive text in eligible MP4 recordings. It does not currently advertise automatic face detection. For manual face masks, use an approved editor with motion tracking; Zight’s web editor can be enabled for supported account workflows.

Frequently asked questions about blurring faces