Hugging Face Daily Papers·· 2026-10-19AI score45
VideoMaMa++: Temporally consistent video matting with preserve and refine tokens
VideoMaMa++: Temporally Consistent Video Matting via Preserve-and-Refine Token
AI briefing
VideoMaMa++ proposes a mask-guided diffusion video matting framework that achieves sliding-window inference through two learnable embeddings, preserve token and refine token, propagates generated mattes across chunk boundaries, and uses a globally shared reference anchor to prevent error accumulation over long sequences, without additional training.
Open original arxiv.org