Skip to main content

Engine Tab

Corrected 19 September 2026: there is no Engine tab. Speech recognition, the interruption choices and the silence timeout are on the agent’s Settings tab on the web (Agents > Configure > Settings); on the phone, speech input is under Edit > Voice > Engines. The provider names below were corrected against docs/user-guide/ (Part 3, section 3.3). The other values on this page were not re-checked.
The Engine Tab handles advanced technical settings that govern how the agent processes speech and manages silence.

ASR Settings (Speech-to-Text)

  • Model: Select the speech-recognition provider: ElevenLabs (the default) or Sarvam for Indian languages.
  • Endpointing: Determines how long the agent waits for the user to stop talking before processing.
  • Min Interrupt Threshold: How loudly/clearly must the user speak to interrupt the agent while it is talking?

Buffer Settings

Configure how sentences are streamed from the LLM to the TTS engine. Optimizing these values is key to reducing “Time to First Word.”

Silence Timeout

How long should the agent wait for user input before saying a “re-engagement” phrase or hanging up?
  • Default: 10 seconds.
  • Prompt hint: “If the user is silent, say: ‘Are you still there? I’m happy to help if you have more questions.’”

Transcription Language

Ensure the Engine’s ASR language matches your agent’s spoken language. For Indian regional languages, Sarvam AI is our recommended engine provider.