Introducing Seed Full Duplex Speech Llm Attentive Listening Robust Interference Suppression Enabling More Natural Interaction
- Source
- seed.bytedance.com
- Author
- ByteDance Seed
- Date
Context
ByteDance Seed's Seeduplex keeps listening while it speaks, using incoming sound and conversational to decide whether to start replying, keep listening, or yield to an interruption. The April 2026 announcement describes combining acoustic and meaning-related information so that a thinking pause need not trigger an answer and background conversation need not interrupt one.
The team reports a 40% reduction in premature responses compared with its previous half-duplex system, in which the assistant handled one turn at a time, and says Seeduplex had rolled out in the Doubao app. The intended improvement is conversational timing: users can hesitate or interrupt without forcing every exchange into rigid alternating turns.
- context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
- articleSeedrealtime Audio Visual Full Duplex Llm Released Toward Omni Modal Natural InteractionByteDance Seed
- articleFrom Speech To Audio Creation Introducing The Seed Audio 1 0 Audio Creation ModelByteDance Seed
articleDuplexSpeechBench-IFEval: Evaluating Implicit Instruction Following in Full-Duplex Voice AgentsPuneet Mathur, Dinesh Manocha
Checking sign-in…
Loading comments…