88Z AI, 9 April — ByteDance's Seed team has released Seeduplex, a full-duplex voice large model now fully deployed on the Doubao App. Using a 'listen-while-speaking' architecture, the model achieves simultaneous listening and speaking, significantly reducing error replies and interruptions while cutting stop-judgment latency. It improves conversation fluency and user satisfaction, marking the first large-scale implementation of full-duplex voice technology.
At a glance
- Seeduplex uses a 'listen-while-speaking' framework, cutting error replies by 50% and latency by 250ms
- The model is fully live on Doubao App, marking the first large-scale deployment of full-duplex voice tech
The update
ByteDance's Seed team has officially launched Seeduplex, a native full-duplex voice large model that is now fully available on the Doubao App. Built on an innovative 'listen-while-speaking' framework, Seeduplex enables true simultaneous listening and speaking interactions. In terms of performance, the model features precise anti-interference capabilities and dynamic stop-judgment mechanisms, reducing the error reply rate by 50% and the interruption rate by 40%, while cutting stop-judgment latency by 250 milliseconds. Evaluation data shows that the new model improves conversation fluency by 12% and increases user satisfaction by 8.34%. This advancement marks the first large-scale implementation of full-duplex voice technology in the industry, delivering a substantial upgrade to intelligent voice interaction experiences.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 字节跳动SeedSource published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.