Today we are releasing JST-2, the world's first full-stack physics-native world foundation model.
Many world models learn in pixel/video space. We went back to first principles and built the entire model stack in 3D.
Generate a 3D scene in seconds and expand it in real time. Create characters with motion built in, all through one general architecture.
Giga-comparison of a ton of character swap options:
Minimax H3 Image: Default, @_akhaliq , and @Alissonerdx swap loras.
Qwen Image 2.1: Alissonerdx BFS Body and @ViggleAI swap loras
Krea 2: ID Edit V2.0 and Allissonerdx BFS swap loras.
Char sheet and lora links in thread.
I used Fable 5.2 to build the same world twice, once with meshes and once with Gaussian splats.
Meshes took 200M tokens, 8 hours and 10 rounds of iterations. Fable modeled everything in code and Blender. The girl was quite rough and didn’t look like her picture. Modeling characters from scratch isn’t there yet.
3DGS took 110M tokens, 3 hours and 100 credits through the @Viggle_PINOC MCP. The worlds and the girl were generated from single images, and her moves came from text to motion and video to mocap.
For this project, I definitely feel the 3DGS one is much closer to the references and keeps the dreamy feel I wanted.
I used Fable 5.2 to build the same world twice, once with meshes and once with Gaussian splats.
Meshes took 200M tokens, 8 hours and 10 rounds of iterations. Fable modeled everything in code and Blender. The girl was quite rough and didn’t look like her picture. Modeling characters from scratch isn’t there yet.
3DGS took 110M tokens, 3 hours and 100 credits through the @Viggle_PINOC MCP. The worlds and the girl were generated from single images, and her moves came from text to motion and video to mocap.
For this project, I definitely feel the 3DGS one is much closer to the references and keeps the dreamy feel I wanted.
I used Fable 5.2 to build an interactive web piece where you walk a girl through eight worlds. Jump through a portal into each new one with a dive, a leap, or a cartwheel.
The girl and scenes are all 3DGS, each generated from a single image.
Breakdown & playable link 🧵
Ultrafast is now live for MiniMax H3 in Viggle ⚡
It makes a 10-second 768p video in just 10 seconds, about 1/5 of the time Standard H3 takes. That's the actual user experience on Viggle, from click to video.
You no longer have to wait 3 mins for a 10-second clip.
Try it at viggle.ai/meme/app/video-gen
World Labs released Atlas today. We shipped JST-2 on Oct 19.
Both let you expand a 3D world one view at a time. Here they are side by side.
Left, JST-2: every expand is native 3D, generated in about a second. Walk through a door and the next room is there.
Right, Atlas: every render is a 2D frame, shown as a point cloud. Turning it into 3D splats is a separate step: 5,000 credits and about 15 minutes.
JST is physics-native. It models persistent atoms in 3D space, so 3D is the output, not a post-process.
World Labs has raised over $1B. Viggle has raised under $30M. Now imagine this architecture at scale.
Try it free: viggle.ai/jst2
I used Fable 5.2 to build an interactive web piece where you walk a girl through eight worlds. Jump through a portal into each new one with a dive, a leap, or a cartwheel.
The girl and scenes are all 3DGS, each generated from a single image.
Breakdown & playable link 🧵
Replaced the entire cast in one go, while keeping every character perfectly consistent across scenes, cuts, and camera angles.
created this trending video with @ViggleAI
Here is how you can do full multi-character scene remakes step-by-step👇🧵
I used Fable 5.2 to build an interactive web piece where you walk a girl through eight worlds. Jump through a portal into each new one with a dive, a leap, or a cartwheel.
The girl and scenes are all 3DGS, each generated from a single image.
Breakdown & playable link 🧵
Reshooting a scene because you want a different hero, a different planet and a different prop is the kind of thing nobody does.
So I tried asking for it instead.
One football clip, two messages, and the footballer is now an astronaut juggling a glowing orb on another world. The real Viggle UI is in the middle of the video.
@ViggleAI
What happens when Sam Altman and Tesla Optimus enter The Wolf of Wall Street?
The result is pure AI cinema madness.
@ViggleAI Remake brings iconic movie scenes to life with AI, preserving the original timing and expressions.
Tutorial below 👇
Viggle is doing something seriously impressive, and I think creators should pay attention! 🔥
What makes Viggle Remake stand out isn't just its ability to transform videos. It's how much creative control it gives you over an existing scene.
Unlike many AI video tools that focus primarily on generating clips from scratch, Viggle lets you take an existing video and fundamentally reinvent it.
Here's what makes it different:
🎬 Full Character Recasting: Replace every character in a scene while maintaining character consistency across multiple cuts.
🎭 Beyond Face Swapping: Change entire characters, not just their faces. Reimagine who appears in your videos.
🌍 Scene Transformation: Swap objects, change backgrounds, restyle environments, and introduce new visual effects.
⏱️ Up to 2-Minute Videos: Work with longer sequences rather than being restricted to short clips.
💬 Conversational Editing: Keep chatting with the AI to correct individual shots, refine results, and get the scene exactly how you want it.
🔓 Open Weights: Viggle is releasing its model weights, giving the AI community more opportunities to explore and build with the technology.
💰 Affordable Creation: Starting at $0.96 per 10 seconds at 720p.
But here's what I find particularly exciting: Viggle is shifting the focus from generating a video to having the power to remake an entire video.
That's a major creative advantage for filmmakers, advertisers, storytellers, and AI content creators.
Big respect to the Viggle team for pushing the boundaries and opening up this technology to the community.
Workflow in the comments
Try it: viggle.ai/meme/app/home
Explore the open weights: huggingface.co/Viggle/Viggle…#Viggle#remake
I have made many tsundere #malfoid content but as a malfoid enjoyoor loving the idea that Harry is begging for more of malfoid 💋
Does any other malfoid enjoyoor likes this clingy Harry?
Made it with @ViggleAI Remake
#malfoy#HarryPotter