MovieGrid uses multi-grid post-training to make long-form multi-shot video generation coherent
Jiawei Mao · hf · 2026-09-09
A new method called MovieGrid decomposes long videos into spatially arranged chunks for joint modeling, improving multi-shot coherence while scaling video length efficiently. Published on Hugging Face by Jiawei Mao, it targets long-form multi-shot video generation via multi-grid post-training.
More from Multimodal
- Mask Forcing curbs mode collapse in autoregressive video diffusion distillation — Zhuoran Zhao · 2026-09-09
- Creator builds custom tool for translucent cyanotype/riso AI animation — floguo · 2026-09-09
- AI Circle Buzzing Over a New Image Model Users Call Addictive — Dimillian · 2026-09-09
- Heavy Grok Bot User Shares First Impressions of Meta Muse After Weeks of Daily Use — Scobleizer · 2026-09-09
- fal Researcher to Host Workshop, Talk on Training Video Models with Three.js at Paris Three.js Conf — OdinLovis · 2026-09-09
- GTA VI pixel-art animation made with MiniMax H3 ref2vid — Time-Ad-7720 · 2026-09-09