MosaiChunk adds persistent visual memory to autoregressive video models
xiuyu_l · x · 2026-10-13
MosaiChunk introduces persistent visual memory for long-horizon autoregressive video generation — instead of remembering everything, it selectively remembers the right things, addressing memory loss across long video sequences.
Related event: MosaiChunk Gives Autoregressive Video Models Persistent Visual Memory(2 posts)→
More from Multimodal
- Keyframe Labs launches persona-2.0-live avatar model with word-level emotion control and 180ms latency — edgarpavlovsky · 2026-10-13
- A24's 'The Bird' documentary by Nathan Fielder blends real and AI-generated footage — AIandDesign · 2026-10-13
- Unicorn prompt showdown: two models, two wildly different art styles — imjustnewatai · 2026-10-13
- MosaiChunk: compositing spatio-temporal KV memory for autoregressive video generation — camenduru · 2026-10-13
- SAM3 inverted-mask + MiniMax H3 workflow swaps characters in video, files shared — LordDarthShader · 2026-10-13
- Workflow: Using Claude Opus 5.5 to edit videos — nail the first 15s, then iterate — dotey · 2026-10-13