Get Zlio on your home screen
One tap, full screen, like an app.
Researchers introduce Future Chunk Prediction, an auxiliary training objective that pushes autoregressive video diffusion models to model how scenes evolve rather than rely on temporal redundancy for local continuity.