1. Agent-Assisted and Manually created LoT Video Generation

GPT6 Astra creates Blender scenes or directly draws keyframe importance maps, which we convert into Level-of-Token layouts for video generation. These demos show how an agent can plan where and when to allocate detail.

tokens across the clip
fewer than a uniformly fine grid
Generation prompt

Negative prompt

Scene or painted keyframes → temporal token layouts