Doubao Real-Time Speech Model 3.0 API Service Officially Launched

Volcano Engine has officially launched the Seeduplex Real-Time Speech Model 3.0 API service and opened invitation testing. As a native full-duplex end-to-end speech model, the model has three major advantages: accurate adherence, anti-interference, and dynamic stop detection. It supports calling tools to complete tasks in real-time conversations, enabling simultaneous listening, speaking, and action. Stop detection latency has been reduced by approximately 250ms, and the proportion of interruptions in complex scenarios has decreased by 40%.