At SIGGRAPH 2026’s Real-Time Live!, the audience became the demo. People pulled out their phones, sent in a picture, and seconds later that picture was a fully textured, rigged, and animated 3D character walking around a shared scene on the big screen. The team behind it — VAST, the makers of Tripo AI — walked away with Best in Show.
The Story
Real-Time Live! is SIGGRAPH’s flagship live-demo showcase. No pre-rendered video, no “results may vary” footnote — everything has to run on stage, in front of a few thousand of the most demanding graphics people on Earth. This year it was chaired by Mk Haley and ran eight demos spanning AI, robotics, rendering, and simulation. The one that won was called “Create Interactive 3D Assets in Seconds!”
The pitch is simple to say and hard to build: hand the system an image, get back a complete 3D asset — mesh, texture, skeleton, and motion — in seconds. Not a gray blob. Not a T-posed statue you still have to rig for an afternoon. A character that already knows how to move. The showstopper was letting the room drive it: attendees submitted prompts from their own phones, and each creation instantly dropped into one shared, animated scene. The presenters called it “a new paradigm for interactive 3D content creation,” and for once the room agreed.
The team reads like a who’s-who of the generative-3D wave: Ying-Tian Liu, Hanxiao Wang, Yuan-Chen Guo, Yumeng Li, Yi-Hua Huang, Yu-Lin Tsai, and Yan-Pei Cao — mostly from VAST, with collaborators at the Chinese Academy of Sciences and the University of Hong Kong. VAST is the company behind Tripo, a name regular Lab readers know well.
Why You Should Care
Here’s the part that matters. For two years, AI 3D has been great at the shape and terrible at everything after it. You’d generate a beautiful mesh and then hit the wall the Lab keeps calling the “dead-statue problem” — no bones, no weights, nothing that moves. Rigging and animation stayed slow, manual, and expensive. That wall is exactly what this demo knocked down, live.
And it didn’t come from nowhere. This is VAST assembling parts it has been shipping in public. UniRig gave it one auto-rigger for humans, animals, and stylized props. Tripo P1.0 made the mesh side fast and topology-aware enough for game engines and robotics in around two seconds. Auto-rig and one-click animation stitched the rest. Real-Time Live! was the moment all of it clicked together into a single, seconds-long loop — in front of witnesses who don’t clap easily.
If you make games, this is the asset backlog evaporating: a napkin sketch or a reference photo becomes a movable extra you can throw into a scene before lunch. If you do previz, virtual production, or interactive installations, the “shared scene the crowd fills up” idea is a template you can steal outright. And it wasn’t the only AI moment on stage — the same session featured a Promptable AI Camera, “Game-Ready Hair in Minutes,” and even a Disney Olaf brought to life as a reinforcement-learning robot. Real-time generative creation is no longer a slide. It’s the show.
Try It / Follow Them
- Use the tech now: the image-to-3D and auto-rig pipeline is live in Tripo Studio — drop in an image, get a mesh, rig it, animate it.
- Go open-source: UniRig is on GitHub with weights, if you’d rather build the rigging step into your own pipeline.
- Watch the demo: the full Real-Time Live! 2026 session is up, and the program page lists every presenter.
- Read the recap: SIGGRAPH’s own blog write-up covers all eight demos.
IK3D Lab Take
We’ve covered every piece of this puzzle as it landed — TripoSplat generating splats, Tripo H3.1 doing film-grade meshes, AniGen turning a photo into a rigged character, SkinTokens treating rigging like a language. What Real-Time Live! proved is that the pieces are now a pipeline, fast enough to run on a stranger’s phone photo while a crowd watches. The gap between “I have an idea” and “it’s a moving 3D character in my scene” just collapsed to the length of a text message. Winning Best in Show against hardened graphics veterans isn’t hype — it’s the field admitting the dead-statue era is over. Now go point your phone at something.
