Audio environment
Overview
Audio environment controls how the agent’s outbound audio sounds to the caller. Two settings apply:
- Background sound plays a looping ambient sound under agent speech.
- Voice filter applies a phone-style filter to the agent’s voice before that mix.
The caller hears the result. The agent does not hear the background sound, and it is not sent into speech recognition.
This is separate from tool call sounds, which play only while a tool runs.
Background sound
Background sound mixes a built-in preset under agent speech for the duration of the conversation. The sound keeps looping when the user interrupts; it does not stop with the agent’s turn.
Presets include office, restaurant, city, typing, and elevator music. In the dashboard, you can preview a preset before you save.
Configure background sound with conversation_config.conversation.background_sound.
Configuration
Update via the dashboard
Update via the CLI
Update via the API
On very poor WebSocket connections, mixing background sound can produce choppy audio.
Parameters
Valid source_id values are office1, office2, restaurant, city, typing, and elevator1 through elevator4.
Voice filter
Voice filter reshapes the agent’s generated voice. It does not equalize or denoise the background sound. The filter runs first; background sound is mixed under the filtered speech.
Voice filter is dashboard-only. It is not part of the public agent API or CLI schema, and it may not be available on all workspaces.
In the same Audio environment panel, set Voice filter to Off or a phone preset, then save.
Limitations
- Built-in presets only. You cannot upload a custom file or point at an arbitrary URL.
- Audio environment controls are hidden when the agent is text-only.
- Background sound is not mixed into WhatsApp voice-message replies. WhatsApp calls still include it.
- Background sound cannot be set per conversation through client overrides;
conversation_config_overridedoes not includebackground_sound. You can still change it on a workflow node.