The Viggle Revolution: Meet JST-2, the AI Model Building Physical Worlds in Real-Time
In a world where most AI models learn about the world through pixels alone, Viggle introduces an entirely different approach. With the launch of the JST-2 (Joint Space-and-Time) model, the company is changing the game by offering a Foundation Model that treats physics as an integral part of space, rather than something that needs to be guessed from images.
What is JST-2, really?
Unlike previous models that analyzed video at the pixel level, JST-2 constructs the world using "atoms" of space and time. Every object or part of it is represented as a point in 3D space (XYZ) with vectors of properties and dynamics. The result? A model that understands how things move, deform, and behave in space, without requiring cumbersome rigging processes or manual skeletal building for animation.
Why does this matter for creators and developers?
The ability of JST-2 to generate entire scenes in about one second (on NVIDIA B200 chips) is a significant breakthrough. Here are some of the key capabilities that are changing how we create content:
- •Interactive world-building: You can create a 3D environment, navigate through it, and expand it from new angles in real-time. The model doesn't just "paint" an image; it builds a space that can be explored.
- •Effortless animation: Characters generated from a single image arrive with built-in movement capabilities. There is no need for a studio, Motion Capture suits, or complex editing processes. The model understands the physics of movement and applies it directly to the character.
- •Spatial consistency: Thanks to the use of Physics-Native Tokens, the model maintains the consistency of characters and objects even when the camera moves or the scene changes.
Data and performance
According to data published by Viggle, the model shows impressive performance in blind tests against existing solutions. In the field of Text-to-Motion, the model demonstrated clear superiority in the ability to generate natural, continuous, and accurate movements, especially in complex motions like dancing or hand gestures. The ability to generate a scene in one second makes this tool practical not only for artists but also for game developers and content creators who need high workflow speeds [1, 5].
Summary and conclusions
Viggle is not just building another video generation tool, but laying the foundation for what they call an "Experience Engine." The combination of high rendering speed and a deep understanding of physics allows creators to stop worrying about technique and start focusing on creation.
For those involved in no-code development, artificial intelligence, or virtual worlds, JST-2 is a tool worth following. It lowers the barriers to entry for creating high-quality 3D content and brings us closer to a future where anyone can build entire worlds with the click of a button.
For more information and to test the model: