Voice Agents That Handle Interrupts - Chintan Agrawal and Daniel Wirjo, AWS
Voice agent quality is an audio engineering problem, not an LLM problem, requiring sub-200ms turn detection
“These are all audio engineering problems. They are not LLM problems because you can have the perfect model, perfect track but the experience still might feel broken if the turn taking is off.”
AWS solutions architects present a framework for voice agent turn-taking, arguing the 200ms human conversational switch threshold makes interrupt detection a critical audio pipeline problem independent of model quality. They demonstrate that identical LLMs produce vastly different user experiences based solely on how quickly the audio layer detects and yields to interruptions. The talk covers three approaches from simple silence detection to custom turn-detection models.