Hi team, so I noticed recently when I use vapi the assistant voice speaks too quickly, too slow, too high pitched, etc sometimes, basically creating unpleasant experience for user. Is there an issue going on? I’m using custom llm with eleven lans voices btw

To address the voice speed issue with ElevenLabs voices, try explicitly setting the

speed

parameter in your assistant's voice configuration. For example:

json
"voice": {
  "provider": "11labs",
  "voiceId": "your-voice-id",
  "speed": 1.0 // Adjust between 0.7 (slower) and 1.2 (faster)
}

Fine-tuning this value can help achieve a more natural cadence for your assistant's speech ([documentation](https://docs.vapi.ai/changelog/2025/3/2)). If the issue persists, consider testing with different speed values to find the optimal setting for your use case. Source: - [Changelog March 2, 2025](https://docs.vapi.ai/changelog/2025/3/2)

Hello, For ElevenLabs voice configuration, you can control various aspects of the speech output using SSML tags or configuration settings. Since you're experiencing inconsistent speech patterns, here are some potential solutions: 1. Use SSML tags to control speech:

<speak>    <prosody rate="medium" pitch="medium">        Your text here    </prosody></speak>

2. Verify your ElevenLabs configuration: - Check stability settings (higher stability = more consistent output) - Adjust similarity settings if using voice cloning - Consider using different voice models if issues persist 3. Common issues to check: - Long responses might affect speech timing - Complex punctuation can impact pacing - Model outputs might need formatting for better speech synthesis 4. For custom LLM integration: - Ensure responses are properly formatted - Add appropriate punctuation - Consider adding natural breaks in the text If issues continue: 1. Test with different ElevenLabs voices 2. Monitor your system prompt's formatting 3. Consider adding specific instructions about speech pacing in your prompts You can also try adding specific SSML controls in your assistant's responses to maintain consistent speech patterns.

ah ok thanks, will try the SSML and other points you mentioned

Hello, Sorry for the inconvenience and thanks for your patience. Don’t hesitate to contact us again let us know if you need further assistance.