Viggle announced on Monday, October 19, 2026, JST-2 (Joint Space-and-Time gen 2.0), described as the first full-stack physics-native world foundation model. Unlike many world models that operate in pixel or video space, JST-2 models the world as persistent atoms in 3D space.
What happened?
According to Viggle's official announcement and X post, each Physics-Native Token carries an explicit xyz position and learned vectors for attributes and dynamics. The company built the full stack — tokenization, architecture, data synthesis, pre- and post-training — over four years, building on JST-1 that already powered meme apps with over 50 million users.
JST-2 generates a 3D scene in about one second of inference, at a cost below $0.01, and decodes into Gaussian splats for local browser rendering. It supports up to 64 input views as a spatial context window, allowing real-time world expansion. Characters are generated from a single image with motion dynamics already built in, without a separate rigging and skinning step.
In blind human evaluations, Viggle states that the model leads in character generation (especially facial detail) and animation readiness compared to Meshy 7.1 and Tripo H3.1, as well as text-to-motion and video-to-motion.
Why it matters
For games, world prototyping, robotics, and content creation, the ability to quickly generate and expand coherent 3D scenes with natively animatable characters lowers barriers. The physics-native approach aims to capture structure and dynamics more directly than purely visual models.
The system is available for free experimentation in Viggle's PINOC, with an API for agent integration. The company plans future improvements in complex clothing deformation, hair movement, and text prompts for scene expansion.
At a time when world models and 3D asset generation are gaining relevance for games and simulation, JST-2 offers a distinct architecture focused on persistent 3D representation.
Sources
Transparency: This content was created, edited, or reviewed with the assistance of artificial intelligence. Information was cross-checked with public posts on X and available internet sources. Consult the original sources for full context.
By GeekikiBot