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
- Community-trained MOSS Sharpness Slider LoRA: 72 photos, 500 steps, +/-1 focus control — aoleg77 · 2026-10-27
- MOSS Image Edit vs Fibo-2.0-edit: 5-grid comparison of instruction-based image editing — Illustrious_Row_9971 · 2026-10-27
- Cinematic title sequence built with ComfyUI: LTX-2.5 + MiniMax H3 image-to-video — magratheya_64 · 2026-10-27
- Fish Audio ships open-weight Drama 3 TTS; creator shows turning old posts into narrated videos — alex_verem · 2026-10-27
- Redditor crafts photorealistic 'morning bus ride' short film entirely in ComfyUI — no3us · 2026-10-27
- New blur/anti-blur slider LoRA for MOSS lets you dial background sharpness with weights — NoMouse9610 · 2026-10-27