Video Foreground Extractor

BiRefNet v2 is specialized for "dichotomous" segmentation, meaning it identifies and separates objects from complex backgrounds with high precision. This makes it particularly effective for e-commerce, green-screen replacement, and 3D compositing. This is achieved through a bilateral reference system that captures both global context and fine local details, enabling pixel-perfect knowledge of the visual environment.