guillem_navarra
03/13/2025, 10:15 AMVapi Support Bot
03/13/2025, 10:16 AMjson
{
"transcriber": {
"provider": "deepgram",
"model": "nova-2",
"language": "es" // Change language code as needed
}
}
2. Language Model (LLM) ðŸ§
- OpenAI models like GPT-4 or Google Gemini work well for multilingual support
- These handle the conversation processing across languages
3. Text-to-Speech (Voice) 🔊
- Azure supports ~400 voices across 140+ languages, including all your required languages
- Other options include ElevenLabs, PlayHT, and Deepgram
- Example for Spanish voice:
json
{
"voice": {
"provider": "azure",
"voiceId": "es-ES-ElviraNeural"
}
}
Here's a complete configuration example:
json
{
"name": "Multilingual Assistant",
"model": {
"provider": "openai",
"model": "gpt-4"
},
"transcriber": {
"provider": "deepgram",
"model": "nova"
},
"voice": {
"provider": "azure",
"voiceId": "es-ES-ElviraNeural" // Change for different languages
}
}
For each language, you'll need to use the appropriate voice ID. For example:
- Spanish: "es-ES-ElviraNeural"
- German: "de-DE-HeddaNeural"
I found that Azure specifically supports Catalan, though you'll need to check their documentation for the exact voice IDs to use.
Hope this helps you get started! 😊
Sources:
- [Vapi Documentation](https://docs.vapi.ai/customization/multilingual)
- [Vapi Documentation](https://docs.vapi.ai/customization/multilingual#transcribers-speech-to-text)
- [Vapi Documentation](https://docs.vapi.ai/customization/multilingual#voice-text-to-speech)Vapi Support Bot
03/13/2025, 10:16 AMguillem_navarra
03/13/2025, 10:37 AM