September 7th news, according to Bloomberg, ByteDance is developing a real-time spatial video generation AI model, which is planned to be released as early as October, with founder Zhang Yiming personally overseeing the development. A source said that Zhang Yiming has been continuously coordinating the company's various business departments, mobilizing AI resources and computing power to advance the project, but the exact release date has not been finalized and may still be adjusted.

It is reported that the model is built upon ByteDance's existing movie-grade video generation AI model Seedance, and can generate interactive virtual worlds for live streaming, short plays, and games. Zhang Yiming positions it as a "flywheel" driving the entire ecosystem. Its capabilities are similar to Google's Genie, allowing users to be placed in highly immersive three-dimensional virtual environments, and connecting ByteDance's self-developed AI models, cloud computing resources, content platforms, and the hardware terminals of XR business Pico.

The model also aims to generate virtual worlds that respond to voice or motion of Pico headset users, and transfer the high-computing task of spatial content generation mainly to the cloud, reducing the computing power requirements and usage barriers of VR devices, thus enabling more cost-effective and simplified hardware terminals. If this model is implemented, the focus of competition in the VR industry may further shift from hardware specifications to AI models, computing infrastructure, and content distribution capabilities.