Barge-in allows users to interrupt a voice agent while it is speaking. Unlike chat interactions, which follow a turn-based conversation model, voice interactions are continuous and conversational. With barge-in enabled, users can speak at any time without waiting for the agent to finish its response, creating a more natural and responsive experience.

​

Enable Barge-In for an AI Agent

1. Create an AI Agent. For instructions, see Creating AI Agent Application.
2. In AI Agent Studio, select Admin Panel from the left navigation pane.
3. Select Voice Settings, and then turn on Enable Voice.
   

Note: You can select a speech profile from the available list. The selected profile is applied during voice playback.

4. Barge-In is enabled automatically when you turn on Enable Voice. Select Save.

Configure Barge-In Settings

The following settings control how the AI Agent responds when a user interrupts.

Sensitivity

Controls how much of the agent's response is protected from interruption. This value is configured as a percentage.

For example, if Sensitivity is set to 20%, the first 20% of the agent's response cannot be interrupted. If the user attempts to interrupt during this protected portion, the agent continues speaking.

Time Prompt

Specifies the minimum amount of time the agent must speak before interruptions are allowed.

For example, if Time Prompt is set to 500 milliseconds, users can interrupt the agent only after it has been speaking for at least 500 milliseconds.

Barge-In Response Time

Specifies how long the user must speak before the agent stops its response after an interruption is detected.

For example, if Barge-In Response Time is set to 1,500 milliseconds, the agent stops speaking after the user has spoken continuously for 1.5 seconds. This helps prevent overlapping speech.

If no value is specified, the setting defaults to null, and the agent continues speaking even when an interruption occurs.