RTFM (Real-Time Frame Model)
RTFM generates video of an interactive 3D world in real time on a single H100, using posed frames as spatial memory.
A learned renderer with no explicit 3D geometry. Each frame carries its camera pose, and 'context juggling' retrieves nearby frames so persistence does not grow with session length. Handles reflections, shadows and lens flare in one model.
- Date
- Thursday, 16 October 2025
- Lab
- World Labs
- Kind
- model
- Access
- research preview
Sources
This record was checked against its sources on 6 October 2026. How we check