§ Application · Modality extension through code
An agent that
speaks animates.
We pair Hominis — a text-only reasoning model — with the ReAgent framework to reach a modality it was never trained on: short, pedagogically structured educational videos. Hominis emits Python; ReAgent renders it in a sandbox.
- 300+
- rendered scenes
- 28
- academic domains
- 20–60s
- per scene
- 0
- fine-tuning steps
01
Selected scenes
A hand-picked walk across six domains.
Hover to preview · click for the full brief and a fullscreen replay.