clip-extractor

Automatically reframe a 16:9 video for 9x16 vertical shorts by using face-tracking to detect subjects and calculate per-frame portrait crops. This pipeline applies Kalman smoothing for stable framing, outputting ready 9x16 or 1:1 clips.

Yes, face-tracking supports split-screen multi-face layouts for vertical video crops. The pipeline auto-detects multiple layout modes including dynamic podcast-like layouts, ensuring all subjects are properly framed within the 9x16 output.

Video reframing outputs ready 9x16 or 1:1 clips that pair directly with Remotion or standard editors for shorts and reels. The Python pipeline handles rendering frame-accurately so you can import the results seamlessly.

The best way to convert horizontal video to a 1:1 square format without manual cropping is using automated face detection and layout classification. The system identifies the 1x1 square layout mode and applies per-frame crop calculation with Kalman smoothing.

Face-tracking video reframing needs Kalman smoothing to stabilize the framing around the detected subject. Without it, per-frame crop calculations based on face detection would produce jittery, unstable outputs during dynamic movements in long-form videos.

No, you do not need to specify the layout mode manually for podcast-like video reframing. The pipeline auto-detects four distinct layout modes, including dynamic podcast-like layouts and split-screen multi-face setups, based on face tracking.