Glossary
Auto-Reframe
Auto-reframe is the automatic conversion of horizontal 16:9 footage into vertical 9:16 by tracking the most important subject — usually a face — and panning the crop to follow it.
The naive approach to vertical conversion is a static center crop, which works only when the subject sits dead-center for the full clip. Real footage has speaker swaps, crowd shots, gameplay panning, and split-screen interviews where a center crop loses half the action.
Auto-reframe samples frames at 5-10 fps, runs face and motion detection per frame, then computes a smoothed crop trajectory across the clip. The smoothing matters — jerky pan transitions feel worse than missing the action, so most pipelines clamp pan velocity to about 100 pixels per second.
AutoClip's auto-reframe runs inside the same Modal worker as the cut and caption stages, so an 8-minute source clip reframes in roughly 30-45 seconds. Two-speaker podcast formats trigger a special split-mode that switches between speakers based on voice activity detection rather than face position alone.
Related Terms
Frequently Asked Questions
Does auto-reframe work with multiple speakers?
Yes, with a tradeoff. Two-speaker mode follows whoever is talking via voice activity detection. Three or more speakers usually look better with a wider crop that captures the group, since pan transitions every 4-5 seconds feel disorienting.
Can I override auto-reframe on specific clips?
Yes. AutoClip exposes a manual crop offset per clip in the dashboard. Useful when the speaker is sharing the screen with a graphic that the auto-crop is hiding.
Put Auto-Reframe to Work
AutoClip handles the full pipeline — viral moment detection, 9:16 reframing, captions, and auto-posting. Start clipping for free.
Get Started Free