PixVerse adds shared worlds to R1
PixVerse has updated its R1 real-time world model so users can enter AI-generated worlds as personalized avatars and shape them together in a continuous shared stream with no session limits.
PixVerse announced on 1 April 2026 a major update to R1, the real-time world model it launched in January. R1 turns a text prompt into a continuous video stream that responds to user input, instead of producing a fixed clip.
The update adds three things:
- Personalized avatars. In a new Avatars tab, users upload one to three photos (front, side and back views) to create a character that represents them in the generated world. PixVerse says avatars respond to movement, positioning and animation.
- Shared worlds. Multiple users now take part through a common livestream. Prompts appear to everyone in a scrolling feed. When the model picks one up, it is rendered into the shared scene with visual effects, and a live chat layer runs alongside.
- No session limits. R1 previously capped individual sessions at five minutes. The shared world now runs continuously.
PixVerse credits three components: an "Omni" native multimodal foundation model that handles text, image, audio and video as one stream; a consistency-aware autoregressive framework for temporal coherence; and an "Instantaneous Response Engine" that it says delivers real-time 1080p generation with millisecond-level latency. The company has published no independent benchmarks. Its description of R1 as the "world's first" real-time world model has not been independently verified.
R1 is free to all PixVerse users for a limited time at realtime.pixverse.ai. The release was part of PixVerse's "Power-Up Week," which followed its V6 model on March 30. The Singapore-based company says it has more than 100 million users and reached unicorn status with a Series C round in March 2026.