This library detects and tracks human faces in real time from webcam video using deep learning models via WebGL acceleration. It outputs raw face position, scale, and rotation data that can be overlaid with 3D content. The library integrates with popular 3D engines like Three.js and Babylon.js for rendering AR face filters and supports multiple simultaneous face tracking. Framework-agnostic by design, it can swap neural network models and runs entirely client-side in the browser.