Viggle unveils JST-2, a physics-native world foundation model that generates explorable 3D worlds in seconds
azed_ai · x · 2026-10-20
Viggle Research has released JST-2 (Joint Space-and-Time, gen 2.0), a physics-native world foundation model.
- Core idea: Most world models learn in pixel space, inferring physical structure indirectly from appearance. JST-2 instead models the world as persistent atoms in 3D space—each atom represents a piece of matter or object part and retains its identity.
- Capabilities: Unifies scenes, characters, and motion in a shared 3D world you can enter and explore, with scenes generated and expanded in seconds.
- Demos: The team showed worlds like a windmill village, snow village, and starship cabin. The model is available to try now, with an API offered.
Related event: Viggle launches JST-2, a physics-native 3D world foundation model(11 posts)→
More from Multimodal
- All of San Francisco's Street View imagery turned into a semantic 3D city model — bilawalsidhu · 2026-10-20
- MiniMax H3 video generation runs on a single RTX 3060 12GB via GGUF quantization — Ok-Inevitable8391 · 2026-10-20
- Runway CEO says AI could 'absolutely' replace human actors: 'It's already happening' — c_valenzuelab · 2026-10-20
- Higgsfield ships Wan 3.0 Prime with faster animation-style switching — CurieuxExplorer · 2026-10-20
- Hybrid AI-plus-physical filmmaking pipeline emerges as the future of film production — DavidmComfort · 2026-10-20
- Tiny Moss-image model generates a coherent 40-person crowd photo — Crazy-Repeat-2006 · 2026-10-20