You can use Azure OpenAI as a model provider in Vapi for GDPR-compliant LLM tasks, but not as a voice (TTS) provider. Vapi supports Azure as a TTS provider (typically via Azure Cognitive Services), but Azure OpenAI integration is for language models only. Here's an example of configuring Azure OpenAI as your model provider in Vapi:

json
{
  "provider": "azure-openai",
  "models": ["gpt-4o-2024-11-20"],
  "region": "eastus",
  "openAIEndpoint": "https://your-resource.openai.azure.com",
  "openAIKey": "your-azure-openai-api-key"
}

For voice, select "azure" as the provider, but this is separate from Azure OpenAI. [See documentation for details](https://docs.vapi.ai/providers/model/azure-openai) Source: - [Azure OpenAI Model Provider Documentation](https://docs.vapi.ai/providers/model/azure-openai)

Message marked as helpful by @Nicolas Eugui! 🎉

Hi Nicolas Eugui, If you are trying to set up Azure OpenAI with Vapi and encounter issues, ensure that: 1. **Correct Endpoint**: Use the base Azure OpenAI endpoint URL, formatted like

https://your-resource.openai.azure.com

, without including the full endpoint paths. 2. **Deployment Names**: Make sure your Azure deployment names exactly match Vapi's supported model names, such as

gpt-4o-2024-11-20

. 3. **Valid API Key**: Verify your Azure API key is correct and corresponds to the resource you are trying to access. 4. **Supported Region**: Check that your Azure deployment is in a supported region like

eastus

,

swedencentral

, etc. 5. **Optional API Management Key**: The

OCP-APIM-Subscription-Key

is optional unless specifically required by your setup. For more detailed troubleshooting, you can consult the [Azure OpenAI Integration Guide](https://docs.vapi.ai/providers/model/azure-openai).

Thanks for the clarification. I understand that Azure OpenAI can be used as an LLM provider. However, in my case what I need is to use Azure OpenAI Realtime TTS as the voice provider. Right now, in Vapi I only see OpenAI Realtime (direct) available for voice, but not the Azure-hosted version. Could you confirm if there is any way to configure Azure OpenAI Realtime for voice inside Vapi, or if this is not supported yet?\

Vapi only supports OpenAI's direct Realtime API for speech-to-speech functionality, not the Azure-hosted version. The docs specifically mention: 1. For OpenAI Realtime: - Only OpenAI voices may be selected while using this model - The model to use is

gpt-4o-realtime-preview-2024-12-17

- The voice selection works within the speech-to-speech model, not as a separate TTS While Azure OpenAI can be used as a model provider for text completion, there is no current documentation indicating support for Azure OpenAI's Realtime TTS service as a voice provider.