like i said fiat uno but it read it as fiat or no. which is not what i want https://cdn.discordapp.com/attachments/1404815835289096293/1404815835545079890/image.png?ex=689c9033&is=689b3eb3&hm=00d345f5ce58ea235f7d70d8d3b8d409eb3366c81632d42a1936d5879a849043&

To improve recognition of phrases like "Fiat Uno," try switching your assistant's speech-to-text provider to a more accurate model or a custom transcriber. For example, you can configure Deepgram as your custom transcriber in your assistant setup:

json
"transcriber": {
  "provider": "custom-transcriber",
  "server": { "url": "wss://your-server.ngrok.io/api/custom-transcriber" }
}

This allows you to leverage advanced transcription services that may better handle brand names and uncommon phrases. [See more in the Vapi documentation.](https://docs.vapi.ai/customization/custom-transcriber) Source: - [Custom transcriber | Vapi Documentation](https://docs.vapi.ai/customization/custom-transcriber)

@Vapi Support Bot its not about the transcribe but also listens the data and passes the data incorrectly to the function tools attached

To help ensure the correct data is passed to your function tools, try refining your assistant's system prompt to explicitly instruct the model on when and how to trigger each tool. For example:

json
"messages": [
  {
    "role": "system",
    "content": "[TASK] Only trigger the function tool if the user's input exactly matches a known brand or model (e.g., 'Fiat Uno'); otherwise, ask for clarification."
  }
]

This can help reduce accidental tool calls from ambiguous or misheard phrases. [See more on customizing tool behavior.](https://docs.vapi.ai/customization/tool-calling-integration#integrating-tools-with-vapi) Source: - [Integrating Tools with Vapi](https://docs.vapi.ai/customization/tool-calling-integration#integrating-tools-with-vapi)

Hi kd, If the VAPI assistant is misunderstanding words during calls, it could be linked to issues in speech recognition accuracy. Here are some steps you can take to improve this: 1. **Transcription Accuracy**: If incorrect transcription is a problem, consider switching to a more accurate speech-to-text service provider. 2. **Check Call Logs**: Review call logs for transcripts and any error messages to identify the source of misunderstandings. More details on troubleshooting can be found [here](https://docs.vapi.ai/debugging). 3. **Model Configuration**: Ensure that the language model being used is appropriate and configured with optimal settings like context window size and allowable token limits. 4. **Test and Debug**: Utilize the dashboard's testing tools to simulate calls and verify how the assistant handles specific inputs. 5. **Voice Quality**: If the assistant sounds robotic or unnatural, consider switching to a different voice provider or tweaking your current voice settings. For systematic troubleshooting, refer to the detailed steps for debugging issues with voice and understanding in our [guide](https://docs.vapi.ai/debugging). These actions should help in diagnosing and improving the assistant's speech recognition performance.