Hi, I hope you are well. I have an assistant with a script hook "Hi, is this sarah from x clinic?!" This assistant sometimes makes some incomprehensible conversations. For example, in this conversation, instead of the customer's name, it makes incomprehensible conversations. https://storage.vapi.ai/8aeec76d-b09e-4ddd-b703-396c586ccd41-1743458754805-ceb34fc6-b002-4525-960b-dd4e9c48b872-mono.wav I don't know why this happens. Please check it.

Hi there! 👋 Sorry to hear you're experiencing issues with incomprehensible conversations in your Vapi assistant. While I can't directly access or listen to the audio file you shared, I can definitely help explain some potential causes and solutions for this type of issue, especially related to script hooks and customer names. ## Potential Causes of Incomprehensible Conversations There are several reasons why your assistant might be producing incomprehensible speech when trying to use the customer's name: 1. **Script Hook Placeholder Issues**: According to the information I found, script hooks use placeholders like

{{variableName}}

to insert dynamic content such as customer names. If these placeholders aren't properly configured, the assistant might try to pronounce the placeholder itself or produce garbled speech. 2. **Customer Name Availability**: The customer name might not be available when the first message is spoken. As noted in [Discord discussions](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1343862805463302184), "The first message is spoken when the customer calls up, and at that point, the customer name might not be available." 3. **Text-to-Speech Pronunciation Issues**: Some text-to-speech voices may struggle to pronounce certain words correctly, leading to incomprehensible speech. According to [Vapi documentation](https://docs.vapi.ai/quickstart/phone/inbound), this is a known issue with some voice models. 4. **Voice Distortion**: Voice distortion can occur due to network issues, encoding/decoding errors, or problems with voice synthesis. If you're using 11Labs voices, a high "style" property (e.g., 0.8) can lead to complications and unpredictable AI behavior, as mentioned in [Discord discussions](https://discord.com/channels/1211482211119796234/1211483291191083018/threads/1336377512233664523).

Hey! To help track down this issue, could you share: - When exactly this happened (the timestamp) - What response you expected to get - What response you actually got instead This would really help us figure out what went wrong!