Wan-Streamer: Treating Video as World + Event Stream
bronzeagepapi · x · 2026-07-18
Alibaba's Wan team has introduced Wan-Streamer v0.3, which conceptualizes video uniformly as a "World + Event Stream".
- World: Relatively stable context such as scenes, characters, audio, and environments.
- Event Stream: Time-varying content like speech, actions, gestures, and sound reactions.
With this definition, the model treats "predicting how the world moves, changes, and responds in real-time given the world and inputs" as a general pre-training task. The paper highlights its focus on real-time full-duplex audio-visual interaction, treating the event stream as the agent's language, behavior, and action output to achieve low-latency synchronized generation. Reported metrics include 640×368 resolution, 25 FPS, a 160ms streaming unit, 200ms model-side response latency, and 550ms total interaction latency (under a 350ms bidirectional network budget).
More from Multimodal
- Magnific teases October 28 release: 'We're making it move' — cuenca · 2026-10-27
- Midjourney --sref demo: retro Americana Batman art series — gizakdag · 2026-10-27
- AI video is a new medium, not cinema or animation — and it's perfect for depicting dreams — Kyrannio · 2026-10-27
- Dad turns his toddler's 31-year-old quote into a song via SUNO kids' choir — NathanpmYoung · 2026-10-27
- Band's local-AI music video hits 200k views: 900 generations kept 64, 3-second clips beat face drift — Hwang0500 · 2026-10-27
- Adobe Firefly video editor walkthrough: one-sentence Quick Cut turns 32s into 20s — LudovicCreator · 2026-10-26