Definition
Barge-In
A caller's ability to interrupt an AI agent mid-speech and redirect the conversation.
Barge-in (also called interrupt handling) is the ability for a caller to interrupt an AI voice agent while it is speaking and have the agent stop, listen to the new input, and respond appropriately.
Without barge-in support, callers must wait for the AI to finish its complete response before speaking — leading to frustrating, unnatural conversations where the caller cannot redirect or correct the agent.
How barge-in works
- The STT engine runs continuously, even while the TTS is outputting audio
- When speech is detected above a threshold volume, the system evaluates whether it is a substantive interruption or a brief affirmation
- Brief affirmations ("uh-huh," "yes," "okay," "right") are typically ignored — the AI continues speaking
- Substantive interruptions ("wait, actually I have a different question") trigger an interrupt: TTS playback stops, the new speech is fully transcribed, and the LLM generates a new response from the updated context
Challenges
- False positives: Background noise triggering unintended interrupts
- Crosstalk: The agent's own TTS audio bleeding back into the microphone and being detected as speech
- Intent disambiguation: Determining whether "yeah" is agreement or the start of a new thought
TurboCall's barge-in system uses acoustic echo cancellation and interrupt confidence scoring to handle these edge cases reliably.