Turn detection
Turn detection is how a voice agent decides that the caller has finished speaking and it is the agent’s turn to reply.
People do not end a turn with a fixed pause. They trail off, pause mid-sentence to think, or finish with a rising tone that invites a reply. A voice agent that waits for a long silence feels slow; one that jumps in at the first pause cuts people off.
Simple systems use voice activity detection plus a silence timer. Better ones also look at the words so far and how they were spoken, for example treating “my number is 98…” as unfinished however long the pause. Getting turn detection right is one of the largest levers on how natural an agent sounds.
