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.
Interactive 3D scene
Drag to orbit · scroll or pinch to zoom. The amber camera and its viewing frustum follow the camera used to rasterize the layout.
3D preview unavailable in this browser. View the rendered scene alongside.
tokens across the clip
fewer than a uniformly fine grid
Generation prompt
Negative prompt