Can I set AutoDetect for Deepgram STT?
# support
a
Hi there, I have a few dual language speaking VAPI agents and I find if I set deepgram to one specific language it has issues when the other is being spoken, for example French Canadian. I know deepgram has multi for english spanish but does not have this feature for other dual languages. Deepgram does have detect_language=true but how can I tell VAPI to do this (https://developers.deepgram.com/docs/language-detection)? Maybe we could have a drop down option in the transcriber area called AUTO Detect that would enable this? Thanks!!
v
@AiTitus :) Hello! I understand you're looking to enable automatic language detection for your dual language speaking VAPI agents, particularly for languages where Deepgram doesn't offer a combined model like it does for English and Spanish. Currently, VAPI does not have an explicit "AUTO Detect" option in the transcriber settings. However, you can achieve automatic language detection by setting the
detect_language
parameter to
true
in your API request to Deepgram. When this parameter is enabled, Deepgram will automatically detect the language being spoken from the list of supported languages. To integrate this feature into VAPI, you would need to modify the API call to include
"detect_language": true
. If you're looking for a more user-friendly solution like a dropdown option in the transcriber area, this would require an update to the VAPI interface. I would recommend reaching out to the VAPI development team to suggest this feature. For now, if you need assistance with modifying your API calls to include language detection, please let me know, and I can guide you through the process. If you have any other questions or need further support, feel free to ask!
You can continue this conversation by mentioning me in the message.
Got it. Tagging @User for additional help.
s
you can use codeSwitchingEnabled to auto switch deepgram language and use 11labs multilingual voice to speak in different languages and lastly ask LLM to generate respons in user langauge and then it will work as your expecting.
Copy code
{
...exisintingAssistantCo
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-2-phonecall",
    "language": "en",
    "codeSwitchingEnabled": true,
  },
  "voice": {
    "fillerInjectionEnabled": true,
    "provider": "11labs",
    "voiceId": "ryan",
    "model": "eleven_turbo_v2_5"
  }
}
a
Thanks so much!