Wan-Streamer v0.2 upgrades the native-streaming audio-visual interaction model from 192×336 to 640×368 output while retaining approximately 200 ms model-side signal-to-signal latency at 25 FPS, according to the paper abstract. The system preserves a single-GPU low-latency thinker for perception, language/state processing, cache construction, and decoding, while scaling the expensive visual latent generation stage across a multi-GPU Ulysses-style context-parallel performer. The authors report approximately 550 ms total remote interaction latency when a 350 ms bidirectional network budget is included.
No heat snapshots are available in the last 24 hours.