AI Face Tracking & Auto Reframe
ClipSonic reads every shot in a clip — who is on screen, where they are, and when the video cuts — then frames each one on its own terms as it crops to vertical.
Auto reframe usually means guessing one crop position and holding it for the whole clip, which lands between two people or cuts off a speaker who moves. AI face tracking follows the subject instead — but the naive version pans across your source's own cuts, which looks wrong.
ClipSonic does both properly and locally. It detects every shot change, picks a one-, two- or three-pane layout per shot based on who is actually in it, tracks each face with a virtual camera that eases rather than snaps, and cuts the framing on your edits instead of sliding across them. The crop is sized from the face, not the frame, so it never reaches the next person along.
Six framing modes, switchable per clip
Auto Frame handles most footage. The rest are there for when you'd rather decide: Face tracking follows one speaker full-frame with headroom; Split duo stacks two people tracked separately; Split + gameplay puts a tracked facecam over your own background video; Center crop is a straight centred crop with no detection; Original keeps your source aspect ratio untouched. Framing is a per-clip choice you can change and re-render any time.
- Output shapes: 9:16, 4:5, 3:4, 1:1 and 16:9, up to 4K
- Headroom framing — subjects sit slightly above centre
- Detection runs on your machine — no upload, no per-clip cost, no queue
What the framing engine does
Shot detection
Finds every cut in your source and reframes each shot separately, so a wide shot and a closeup are each framed right.
Adaptive layout
One, two or three panes, chosen per shot by how many people are on screen — with the lead taking the big pane in a three-way.
Face-sized crops
Each crop is sized from the subject, close enough to frame them and never wide enough to catch a neighbour.
Smooth virtual camera
Eases in and out with a dead zone, so small head movements don't move the frame and cuts never pan across a source edit.
Re-identification
Someone who turns away or leaves the shot is still the same person when they come back.
Gameplay + facecam
A dedicated split that tracks your facecam over a looping gameplay background.
Common questions
What is AI face tracking in video?
The software detects faces frame by frame and moves a virtual crop to keep the subject in the shot as they move. ClipSonic uses it to reframe horizontal video to vertical without losing the speaker.
How is this different from a plain center crop?
A center crop is fixed and often lands between two people or off the speaker. ClipSonic sizes and positions the crop from the tracked face, and changes it per shot.
Does it handle more than one person?
Yes — one, two and three-person layouts, chosen automatically per shot, with each face tracked independently.
Will the camera pan across cuts in my video?
No. Shot detection finds every edit in the source and the framing cuts with them instead of sliding across.
Does face detection upload my video?
No — detection and reframing run locally on your machine.
Can I change the framing after it is applied?
Yes, per clip — pick a different mode and re-render just that clip.
Related pages
See the reframe on your own footage
Download ClipSonic free and auto-reframe a clip to 9:16 with face tracking — rendered locally.