Independent guide · Not affiliated with Viggle
Meet Viggle’s world foundation model. Understand the 3D scenes, characters, and motion behind the October 2026 release.
The model, explained
JST-2 puts scenes, characters, and motion into a shared 3D world.
Viggle calls it a physics-native world foundation model. The useful distinction is what the system represents: spatial elements that persist across viewpoints and time, rather than only the visible pixels of a finished video.
Underneath are persistent atoms, or Physics-Native Tokens, carrying position, attributes, and learned dynamics. The model decodes these into 3D Gaussian splats, a representation used to render a scene from different viewpoints.
For creators, the release connects environment building with character creation and full-body motion. This guide separates those announced capabilities from the practical entry point, PINOC, and the physics limitations that remain.
Read Viggle’s technical introductionJST-2 supports up to 64 input views for scene inference. A shared spatial representation lets the system work with a scene from multiple angles, rather than treating every view as an unrelated image.
Viggle reports about one second and below $0.01 per scene inference on NVIDIA B200 hardware, excluding network latency. These are model inference figures, not end-to-end speed guarantees or consumer pricing.
The release describes characters created from images or text, with motion built into the model. A separate rigging stage is not required for that workflow. Start by assessing the character’s shape and proportions before asking for more complicated movement.
Text or video can guide full-body motion. A useful first experiment is a single, readable action. Inspect foot contact, balance, and changes in body shape before judging how well an output fits your intended scene.
From reading to exploring
PINOC is Viggle’s official browser entry point. Generation happens there, not on this website.
Visit the official PINOC app. Follow its current account and access instructions. Check the available limits and terms before uploading anything.
Start with Scene for an environment, or Agent to bring scenes, characters, and motion together. Keep the first experiment focused so it is easier to understand the result.
Describe one subject and a clear setting or action. Examine the output from the available views, then change one detail at a time.
JST-2, without the guesswork.
JST-2 is Viggle’s physics-native world foundation model, announced on October 19, 2026. JST stands for Joint Space-and-Time. It brings scenes, characters, and motion into a shared 3D representation, using persistent atoms called Physics-Native Tokens. These carry position, attributes, and learned dynamics, and are decoded into 3D Gaussian splats.
No. This is an independent editorial guide, not an official Viggle website or a generation service. The model, PINOC product, and release imagery belong to their respective owners. All links for trying the model take you to Viggle. This site does not provide accounts, accept payments, or process prompts.
Viggle presents PINOC as a free browser entry point. Account requirements, usage limits, available features, and current terms are determined by Viggle. Its reported sub-cent scene inference cost describes model computation on NVIDIA B200 hardware, excluding network latency. It is not a retail price, an API quote, or a promise of unlimited free generation.
Viggle has an official developer platform and API documentation. Check those sources for the endpoints currently available, authentication, usage terms, and pricing. A capability described in the JST-2 research announcement should not be assumed to have a publicly available API endpoint.
The central distinction is representation. A video model produces a sequence of frames; JST-2 represents a scene and its contents in 3D, with motion in the same system. Persistent spatial elements give the model a way to describe a world across viewpoints and time. This does not mean every interaction is physically simulated or that every output will be consistent.
In JST-2, scene collisions and object interactions still use rule-based systems. Viggle describes learned physics for JST-3 as future research. That direction is not a current JST-2 feature, and the announcement should not be read as a release or availability promise for JST-3.
Based on Viggle’s release. Sources reviewed .