Undocumented Subject Action Detection And Tracking For Frame Aware Cropping

via “face detection and speaker tracking across video frames”

A python tool that uses GPT-4, FFmpeg, and OpenCV to automatically analyze videos, extract the most interesting sections, and crop them for an improved viewing experience.

Unique: Combines face detection with temporal tracking to build a continuous spatial map of speaker positions, enabling intelligent cropping that maintains focus rather than static frame selection. Uses OpenCV's optimized detection pipeline for real-time performance on CPU.

vs others: More intelligent than fixed-aspect cropping because it adapts to speaker position dynamically, and faster than ML-based attention models because it uses lightweight Haar Cascade detection rather than deep learning inference on every frame.