Multilingual transcription
# support
n
Hi, in the roadmap, it says that VAPI now support multilingual assistants https://roadmap.vapi.ai/feature-requests/p/multilingual-bot-for-inbound-calls-auto-detected. Here it says that whisper can be used for multilingual transcription. In the documentation it's not clear how to set the transcription mode to multilingual (it only lists a set of languages). Could you let me know how?
v
Hi @nisalr, I'm an AI assistant for Vapi. I'll analyze available sources and get back to you shortly.
----------------------------- **While I generate an answer, here are some sources that might be helpful**: * [Introduction]() * [Web Calling]() * [Dashboard]() * [Inbound Calling]() -----------------------------
### Answer To set up multilingual support in Vapi using the Whisper model from Talkscriber for transcription, you will need to configure the transcription model in your assistant's settings. Here's a general guide on how to proceed: 1. **Choose the Transcriber**: When setting up or modifying your voice assistant, specify 'Talkscriber' as your transcriber provider and select the 'Whisper' model. 2. **Set the Language**: Even though Vapi supports multilingual functionality, you still need to choose a primary language from the supported list. This is the language your assistant will default to, but can switch from based on the conversation context. 3. **Configure the Assistant**: Ensure your assistant's settings in the Vapi dashboard are adjusted to handle multilingual capabilities, ensuring that it can dynamically switch languages during a conversation if necessary. Here’s an example JSON configuration snippet that you might use when setting up or updating your assistant via the Vapi API:
Copy code
json
{
  "transcriber": {
    "provider": "Talkscriber",
    "model": "Whisper"
  },
  "voice": {
    "provider": "PlayHT",
    "voiceId": "en-US-Jennifer"
  }
}
This configuration sets the Whisper model as the transcriber and uses 'Jennifer' from PlayHT for English text-to-speech. Adjust the
voiceId
and possibly the
provider
to fit the specific requirements for your application's typical user base or targeted languages. For more detailed information on setting up transcribers and other components of your assistant, refer to the Vapi documentation for [Assistants]() and [Provider Keys]() [(1)]() . If the documentation feels unclear or incomplete regarding the multilingual setup using Whisper, it may be helpful to check for updates or clarifications in the latest documentation revisions or contact Vapi’s technical support for more personalized assistance.
s
Basically, we marked it as finished because it was no longer necessary. Most of the providers have their own native multilingual capabilities, so you can use them directly. Therefore, there was no need for us to create a new multilingual model. For now, if you want to have multilingual part, simply create one assistant for call forwarding, and assistants for each of your supported languages. Just ask the user which language they want to speak in, and forward the call to the respective language assistant.
n
okay got it. Thanks!
4 Views