ByteDance Develops Real-Time Spatial Video Model
ByteDance is preparing an artificial intelligence model for real-time spatial video generation, according to a Bloomberg report published September 7. Founder Zhang Yiming is personally overseeing the project, sources familiar with the matter told Bloomberg.
The reported project could enter the market as early as October. However, the launch schedule remains uncertain, and ByteDance could change its plans.
The move would place ByteDance in the growing race to develop AI-powered “world models.” These systems aim to create virtual environments that respond to users and changing inputs in real time. Consequently, the technology could have applications across gaming, extended reality, robotics and autonomous systems.
Model Builds On Seedance Technology
The new system reportedly builds on ByteDance’s Seedance video-generation technology. Seedance already focuses on producing cinematic video content. However, the planned model would extend that capability toward interactive three-dimensional environments.
According to the reported plans, users could create virtual worlds for livestreams, short-form dramas and games. Moreover, the environments could respond to users’ voices and movements through Pico headsets.
The system could reportedly generate content at about 20 frames per second. At the same time, its target latency stands at roughly 0.05 seconds. This would allow virtual environments to react with relatively little delay.
Importantly, ByteDance reportedly plans to handle much of the demanding processing in the cloud. As a result, compatible headsets could require less local computing power. That approach could also help reduce hardware requirements and costs for future immersive devices.
ByteDance Enters World Model Race
The project reflects ByteDance’s broader push into world models. Earlier reporting indicated that the company had made world models one of its major AI priorities for 2026. The company reportedly aims to develop a system that can compete with Google’s Genie technology.
Meanwhile, ByteDance has a significant advantage in AI-generated video. Its Seedance technology provides an existing foundation for developing systems that understand and generate dynamic visual environments. Consequently, the company could connect its video expertise with its Pico extended-reality hardware.
The strategy also places ByteDance alongside major technology companies pursuing similar technologies. Google and Meta have invested in AI systems designed to understand and generate interactive environments. These technologies could also eventually support robotics and autonomous systems beyond entertainment.
However, ByteDance has not publicly confirmed the reported launch date or the full specifications. Bloomberg’s report relies on people familiar with the project, while other outlets noted that the plans could still change. A ByteDance spokesperson also did not respond to Bloomberg’s request for comment, according to The Next Web.
If launched, the system would mark a significant expansion of ByteDance’s AI ambitions. More importantly, it could connect AI-generated video, cloud computing and immersive hardware within one ecosystem.
