pi-H3, a flow-matching video model built on MiniMax-H3, trends on Hugging Face
Lakonik · hf · 2026-11-22
Lakonik's pi-H3 is trending on Hugging Face as an image-text-to-video model. Tags include flow-matching and distillation, with diffusers and ComfyUI support and an linked arXiv paper; the base model is MiniMaxAI/MiniMax-H3. It's a community effort at video/audio generation with distilled acceleration on top of MiniMax-H3 — notable for multimodal practitioners.
Related event: pi-H3 open-sources 2-step video generation, 5s HD in 3.8s on B200(3 posts)→
More from Multimodal
- GPT 6.2 Astra composes Beethoven-style scherzo, with all 604 key strikes from the score — imjustnewatai · 2026-11-22
- Musk demos real-time AI video generation on half of one AI4 TRIP accelerator — jamesdouma · 2026-11-22
- RAEA engine lets AI agents create videos, 3D scenes and games with React — Promptmethus · 2026-11-22
- Do adaptive optimizers still work? Reddit users doubt it for flow-matching models — Inevitable_Pen9043 · 2026-11-22
- SeaDream 5.5 image demo stuns with 2K photorealistic quality — gen_ericai · 2026-11-22
- MiniMax H3 'Underwater World' clip hailed as one of the best AI videos yet — Cheap_Credit_3957 · 2026-11-22