preprint · arXiv (2026) · Jun 23, 2026
Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models
Lianghua Huang, Zhi-Fan Wu, Wei Wang, Yupeng Shi, Mengyang Feng, Junjie He, Chen-Wei Xie, Yu Liu, Jingren Zhou
In plain English
Introduces a single end-to-end model that places user and agent text, audio and video on one causal timeline, so a digital human can listen, respond and show visible listening behavior in real time. The first version is limited to low-resolution (192p) output. Author list shortened.
real-time avatars
digital humans
conversational ai
latency