On July 17, Ali Tongyi Laboratory announced that the newly released Wan-Streamer v0.2 uses a native streaming architecture and distributed inference topology to simultaneously complete understanding and generation in a single model interaction cycle, making AI's communication methods more close to real people. As an end-to-end full-modal understanding and generation model for real-time duplex interaction, WAN-Streamer's end-to-end interaction delay is around 0.55s in terms of response delay.

On July 17, Ali Tongyi Laboratory announced that the newly released Wan-Streamer v0.2 uses a native streaming architecture and distributed inference topology to simultaneously complete understanding and generation in a single model interaction cycle, making AI's communication methods more close to real people. As an end-to-end full-modal understanding and generation model for real-time duplex interaction, WAN-Streamer's end-to-end interaction delay is around 0.55s in terms of response delay.