VideoMaMa++: Diffusion-Based Video Matting With Temporal Consistency
adobe · hf · 2026-10-21
VideoMaMa++, from Adobe Research, KAIST AI and Korea University, is a diffusion-based video matting framework. It uses separate preserve-and-refine token streams with learnable embeddings to propagate mattes across chunks, delivering temporally consistent, detail-accurate results on long videos. Project page: cvlab-kaist.github.io/VideoMaMa-PlusPlus.
More from Multimodal
- MiniMax H3 I2V: how to stop secondary faces' eyes drifting in 9:16 clips — Downtown-Term-5254 · 2026-10-21
- Mistral's Voxtral-mini-tts-3-0 debuts at #6 of 20 on Voice Arena Hindi TTS board — CodeByPoonam · 2026-10-21
- Offline CLIP image search for manufacturing defect photos: full pipeline walkthrough — OkShirt9372 · 2026-10-21
- Reddit asks for best open-weights image editing model for upscaling and reframing — subnohmal · 2026-10-21
- Wan 3.0 lands on Higgsfield at roughly 1/4.5 the price of Seedance 2.5 — sven_ai · 2026-10-21
- User seeks tools to turn a story into 40-120 coherent, framing-aware image prompts — Malina_Fox · 2026-10-21