ByteDance secretly building real-time world model on Seedance, October launch rumored
dotey · x · 2026-09-09
Per Bloomberg, ByteDance—personally spearheaded by Zhang Yiming—is secretly developing a real-time spatial video/world model, reportedly debuting as soon as October to rival Google's Genie.
- Tech base: Built on the Seedance video model with major compute devoted; core breakthrough is real-time interaction—generating dynamic, interactive 3D virtual worlds from users' voice and motions.
- Cloud streaming: Core compute runs in the cloud, cutting local compute and hardware costs for Pico headsets.
- Numbers: 20 fps cloud rendering, end-to-end latency around 0.05s.
- The thread frames this as another case of "Google bleeds first, everyone else crosses the river" (Transformer→ChatGPT, LaMDA→GPT, Veo 3). Genie is in closed testing; official news may arrive soon.
Related event: ByteDance Secretly Builds Real-Time World Model Under Zhang Yiming(4 posts)→
More from Embodied
- Astra drives a SO-101 arm pick-and-place in MuJoCo sim with zero policy training — shekitup · 2026-09-09
- Robotics researcher argues GPT-6 understands the physical world; Chris Paxton notes the cycle — chris_j_paxton · 2026-09-09
- RhodaAI shares robot manipulation demos in first update since launch — chris_j_paxton · 2026-09-09
- Researcher: LLMs' best use may be generating robot pretraining data — igilitschenski · 2026-09-09
- VC: I'd pay 3x fare for a robotaxi that drives a little less conservatively — saranormous · 2026-09-09
- Physical Prompting Works: Robot Model S1 Follows 4 Different Video Prompts In One Kitchen — alex_verem · 2026-09-09