Skip to main content

Configuring Voice and Interaction Settings for AI Agents

Learn how to configure the voice, language, responsiveness, and call behavior of your voice AI agent.

Maria Cairns avatar
Written by Maria Cairns
Updated over a week ago

Why it matters

Voice and interaction settings determine how your agent sounds, how quickly it responds, how interruptions work, and how natural or task-focused its behavior feels. Correct configuration creates a smoother customer experience and ensures the agent can complete tasks reliably.

Key Concepts

Language mode: Controls whether the agent uses a single language or supports multiple languages in one call.

Speech-to-text mode: Sets how the agent interprets audio, balancing speed or accuracy.

Voice selection: Determines how the agent sounds, including tone, speed, and consistency.

Responsiveness: Controls how quickly the agent reacts to the user.

Interruption sensitivity: Defines how easily the user can interrupt the agent while it is speaking.

Backchanneling: Enables optional conversational fillers like “yeah” or “uh-huh.”

Voicemail action: Defines what the agent does when it detects voicemail.

Duration limits: Caps the maximum length of calls and chat sessions.

Step-by-Step: Configure Voice and Interaction Settings

  1. Open your agent and go to the Editor tab.

  2. Scroll to the Voice and interaction section.

  3. Choose a language mode and select the primary language.

  4. Adjust speech-to-text mode and denoising if needed.

  5. Select a voice and configure its speed, stability, and expressiveness.

  6. Set responsiveness and interruption sensitivity values.

  7. Enable or disable backchanneling and adjust its frequency.

  8. Configure reminders and pacing controls.

  9. Set maximum call duration and optional ambient sound.

  10. Add boosted keywords or pronunciation entries if needed.

  11. Choose a voicemail action and configure the static or dynamic message.

  12. Click Save to apply all settings.

Tips and Best Practices

Language and Transcription

  • Use single language when all callers speak a known language; use multi-language only when needed.

  • Set speech-to-text to accuracy mode for detailed or legal conversations.

  • Enable denoising if callers frequently speak in noisy environments.

Voice Selection

  • Choose a voice that matches the intended purpose.

    • Support or service: calm, steady voices.

    • Sales or outreach: slightly more expressive.

  • Set voice temperature low (0–0.4) for consistent, predictable delivery.

  • Set voice temperature higher (0.5–1.0) if you want more natural tone variation.

  • Keep voice speed between 0.9 and 1.1 for the most human-like pacing.

Responsiveness and Interaction Style

  • Use higher responsiveness (0.7–1.0) if you want faster reactions during troubleshooting or Q&A calls.

  • Use lower responsiveness (0.3–0.6) if you want the agent to wait more naturally between turns.

  • Set interruption sensitivity high when customers frequently cut in.

  • Set interruption sensitivity low when the agent must finish key instructions before stopping.

Backchanneling

  • Turn backchanneling on if you want a more human, conversational feel.

  • Keep frequency low for professional or formal calls.

  • Higher frequency can make the agent feel more casual, but avoid using the maximum unless intentionally designed.

Reminders, Pacing, and Timing

  • Use reminders sparingly. One or two is usually enough.

  • Increase begin message delay if you want the agent to sound more polite and less robotic at the start of calls.

Duration and Ambient Sound

  • Set reasonable duration limits to prevent calls from running longer than intended.

  • Use ambient sound only if it enhances realism; avoid it for formal or sensitive conversations.

Boosted Keywords and Pronunciation

  • Add boosted keywords for local names, product terms, or industry-specific vocabulary.

  • Use pronunciation guides for brand names or uncommon words to avoid mispronunciation.

Voicemail Behavior

  • Use static messages for compliance or standard scripts.

  • Use dynamic AI-generated messages when you need personalized or variable voicemail content.

Troubleshooting

Issue

Possible Cause

Fix

Agent misunderstands user speech

Speech-to-text is set to speed mode or no denoising

Switch to accuracy mode and enable denoising

Voice sounds unnatural or robotic

Voice temperature too low or speed too high

Increase temperature slightly or reduce playback speed

Agent interrupts user too often

Interruption sensitivity too high

Lower interruption sensitivity to slow down cutoff behavior

Agent rarely reacts to user

Responsiveness too low

Increase responsiveness value

Voicemail behavior incorrect

Voicemail action not configured

Set voicemail action to Static or Dynamic and save

Did this answer your question?