How to blur a face in CapCut

The overlay, mosaic and mask method

This works on both the mobile app and CapCut on desktop — the tool names are the same, only the layout differs.

1. Add your clip, then duplicate it onto the layer above. Import the same video a second time and place it directly on top of the original, lined up to start at the same point.
2. Apply Mosaic or Blur to the overlay clip only. Select the top layer, open Effects → Style, and add Mosaic for a pixelated look or Blur for a softer one. This covers the whole overlay clip, not just the face.
3. Mask it down to just the face. With the overlay clip selected, open the Mask tool and pick Circle or Rectangle. Drag and resize it so only the face sits inside — everywhere else on the overlay becomes transparent again, showing the sharp video underneath.
4. Feather the mask edge. Nudge the feather slider up so the covered area blends into the frame instead of sitting on it like a sticker.
5. Keyframe the mask to follow the face. Scrub to each point where the person moves, drag the mask to the new position, and set a keyframe. CapCut interpolates between them, but fast movement needs a new keyframe every second or so to stay centred.
6. Repeat the whole setup for every other face. Each one needs its own overlay clip, effect, mask and keyframe track — they don't share settings.
7. Export at your target resolution once every face is covered for the full clip.

Where this gets slow

The mask-and-keyframe method works, but it scales badly. A person standing still for a two-second clip is a couple of minutes of setup. A person walking across the frame means a keyframe every second or two for the whole shot, checked by eye each time against the preview.

Multiple faces multiply that directly — there's no way to detect a second or third face and apply the same treatment automatically, so it's a fresh overlay, mask and keyframe track per person, and they all need re-checking if you trim or re-time the clip afterwards.

The other cost is precision. Dragging a small circular mask around a moving head on a phone screen is fiddly, and it's easy to leave a sliver of jaw or ear uncovered on a fast pan that only shows up when you review the export at full size.

Editing later compounds it further. Trim the front of the clip, change the pacing, or swap in a different take, and every keyframe you set shifts out of alignment with the footage — there's no way to lock the mask to the face itself, only to a point in time, so a re-edit after the blur is in place usually means redoing the keyframes rather than just nudging them.

What tends to go wrong

A few things trip people up specifically with CapCut's mask-and-keyframe approach:

• Mask drift on fast pans. A whip pan or quick head-turn can outrun a keyframe interval, leaving a frame or two where the face is briefly visible.
• Feather too soft or too hard. Too little feather looks like a sticker; too much lets identifiable detail bleed through around the edge.
• Forgetting the overlay's audio. Duplicating a clip for the overlay also duplicates its audio track — mute it, or the sound doubles up.
• Re-timing after the blur is done. Any trim or speed change to the base clip shifts the mask out of sync, since the mask is keyed to time, not to the face.

CapCut vs. Deface

The faster way

Deface skips the overlay-mask-keyframe chain entirely. Import the same photo or video, and it detects every face on device in about half a second, tracking each one automatically as it moves through the shot — no manual keyframing.

Pick a style — blur, pixelate, a solid block or an emoji — and it applies to every detected face at once. Missed a face, or want to leave one visible? Each face has its own toggle, and you can pixelate a face or draw a manual region over anything the detector doesn't catch, like a licence plate. Export strips the metadata too, which CapCut doesn't touch.