Report
MosaiChunk aims to give long-horizon video models persistent visual memory
The introductory post says it remembers “the right things” instead of everything during video generation.
TLDR
A post introducing MosaiChunk describes it as persistent visual memory for long-horizon autoregressive video generation. Its author says the approach is to remember “the right things” rather than everything.
Combined views
—
1 Source, first seen ago
