Shengshu releases Vidu S1 real-time video model

Shengshu Technology has released Vidu S1, a model built for real-time interactive video. The company describes the release as a move in AI video from "generating a piece of content" toward "continuous interaction."

Founder Zhu Jun announced the model on July 3 at the AI Integration and Application Forum, part of the 2026 Global Digital Economy Conference. His keynote was titled "General World Models: A New Paradigm for Unifying the Digital and Physical Worlds."

According to the company, Vidu S1 supports real-time video calls and voice-controlled video generation. Users can direct a digital character's behaviour by voice, and Shengshu says sessions can run for an unlimited length of time. The model outputs 540p (960x540) at 25 frames per second, rising to as much as 42 FPS. It can build a dedicated interactive character, with a personalised voice, from a single starting image of a real person, an anime character or a pet.

The announcement gives no benchmarks or latency figures, and no independent evaluations. It also leaves out pricing, availability and API access, and says nothing about the model's architecture or size. It is unclear whether Vidu S1 is open to users now or is still a demonstration.

At the same conference, the Beijing Software and Information Services Association named Shengshu a "New Model, New Application Benchmark Enterprise" in its 2025 Beijing digital economy evaluation report.