Smart Face Tracking for Perfect Shorts

AutoShorts detects faces in every frame and picks one subject to follow for the shot, holding onto it rather than flicking between people. The crop then pans smoothly to keep that person in frame in vertical format. It reads the picture, not the audio — so it follows the person it is tracking rather than deciding who is speaking.

It detects every face in the frame and holds one subject for the shot, using a stickiness rule so the crop does not flicker between people. It follows a face, not the audio — nothing listens to decide who is talking.

When no face is detected, AutoShorts uses an intelligent fallback: it fits the video to width and adds a blurred background to fill the vertical space.

Framing is smoothed over a 60-frame window, so the camera drifts rather than snapping. When no face is found, person detection takes over, and busy multi-person shots fall back to a wider layout instead of a bad crop.

Absolutely! Podcasts are one of our most popular use cases. The face tracking handles both single-host and multi-guest podcast formats.