ByteDance Launches Full-Duplex Voice Model Seeduplex for More Natural Conversations

ByteDance recently launched Seeduplex, a native full-duplex voice large model. Compared to the previous half-duplex end-to-end voice model in Doubao, it uses a new 'listen while speaking' framework, significantly enhancing the naturalness and smoothness of interactions.

Seeduplex achieves two key breakthroughs: precise anti-interference capability ignores background noise, reducing error response and interruption rates by half compared to half-duplex models; dynamic pause judgment combines voice and semantic features, cutting interruption rates by 40% for more natural conversation rhythm.

Currently, Seeduplex is fully available in the Doubao app. Users can update to the latest version and select 'make a call' in the chat box to experience the upgraded voice call feature.