SIGGRAPH 2026's Real-Time Live Turns Photos Into Rigged 3D Assets in Seconds

SIGGRAPH's Real-Time Live showcase ran on 21 July at the Los Angeles Convention Center, and Best in Show went to a system that builds a textured, rigged and animated 3D asset from a single audience submitted photo, live, on stage. The team behind it, Ying-Tian Liu, Yuan-Chen Guo, Yumeng Li, Yu-Lin Tsai and Yan-Pei Cao from VAST, with collaborators from the Institute of Automation at the Chinese Academy of Sciences, VAST, and the University of Hong Kong, took an image from the crowd and had a usable asset on screen within seconds rather than the hours a studio pipeline normally needs.

Real-Time Live has run at SIGGRAPH for over two decades as the venue for demos that are not allowed to be pre-rendered or faked. Judges and the audience watch the input go in and the result come out, live, which is why a Best in Show win there tends to matter more to working 3D artists than a splashy trailer. This year's Audience Choice award went to a separate project, Dissectible Anatomy: Embodied Exploration for Education, from Tim McGraw and Jack Myers, which lets students pull apart a virtual body layer by layer.

What makes the winning demo notable for anyone working in games, product visualisation or virtual production is the collapse of a step that has always been slow and skilled: turning a reference photo into a clean, riggable mesh. That work still needs a human to check topology, fix seams and get materials production ready, but a rough asset in seconds changes what is worth prototyping before committing real modelling time to it.

None of this is shipping software yet. What SIGGRAPH rewarded was a research pipeline demonstrated live, and the usual gap between a conference stage and a tool you can install should be expected. Even so, image to 3D generation has been improving fast enough over the last two years that a result like this is a reasonable signal of where asset creation tools are heading next.