https://vapi.ai/ logo
Join Discord
Powered by
Multilingual Transcriber Not Switching to Language...
# support
a
We used the codeSwitchingEnabled feature but the transcriber doest switch to German or Chinese when those languages are spoken. What is the expected behavior here? Does "codeSwitchingEnabled": true allow switching between several languages or just two languages ? Transcriber "model": "nova-2", "language": "multi", "provider": "deepgram", "codeSwitchingEnabled": true
v
message has been deleted
@Achal The 
codeSwitchingEnabled
feature, when set to true, is designed to allow the transcriber to switch between multiple languages during transcription. This means that if the audio contains different languages, the transcriber should be able to recognize and transcribe each language accordingly. However, the effectiveness of language switching can depend on several factors, such as the clarity of the speech, the distinctiveness of each language spoken, and the capabilities of the specific model being used. In the case of the "nova-2" model from Deepgram with the "multi" language setting, it should theoretically be able to handle code-switching between several languages, including German and Chinese, as long as those languages are supported by the model. If you're experiencing issues with the transcriber not switching correctly between German or Chinese, it could be due to limitations in the model's language support or an issue with the feature itself. I recommend checking the documentation for the "nova-2" model on Deepgram's platform to ensure that both German and Chinese are supported languages for code-switching. If the documentation confirms support and you're still facing issues, it may be necessary to reach out to Deepgram's support for further assistance. Would you like me to guide you on how to check the supported languages for the "nova-2" model on your dashboard, or do you need assistance with anything else?
You can continue this conversation by mentioning me in the message.
a
Yes, how do I check the supported languages for the "nova-2" model on your dashboard, or do you need assistance with anything else?
v
@Achal To check the supported languages for the "nova-2" model on your dashboard, please follow these steps: 1. Log in to your VAPI dashboard. 2. Navigate to the 'Assistant' page. 3. Click on the 'Transcribers' tab. 4. Look for the "nova-2" model under the provider "Deepgram". 5. Click on the model to view its details. In the details, you should be able to see a list of languages that the "nova-2" model supports, including whether it supports code-switching for languages like German and Chinese. If you have any difficulty finding this information or if you need further assistance with anything else, please let me know!
s
@Achal could you share the call id?
a
Here are some call ids where Chinese was attempted and failed. Chiense workend when langauge was set to "zh" but when languaget was set to multi the code switching did not work with Chinese 3893df89-89fd-4fa8-992f-216834511536 2fbed560-ef8e-400d-9c65-67fbe62c3900 db40e11a-4224-4f27-a7ac-156341cf3318 db40e11a-4224-4f27-a7ac-156341cf3318
s
@Achal it doesn't works with language 
multi
because 
multi
only works with english and spanish. You can try out the calls without 
multi
model.
a
OK, Is there a way for the agent to speak three languages? Spanish, English and Chinese ? Using Code Switching ?
s
a
📝 Support Request: Code Switching + Transcriber Not Saving Properly (Vapi Dashboard) Hi Vapi Team, I'm encountering an issue with my assistant configuration while using the Vapi Dashboard: Transcriber Provider: Deepgram Model: nova-2 Language: multi codeSwitchingEnabled: true (intended) Voice Model: ElevenLabs Turbo 2.5 I’m trying to support both English and Spanish using the multi language setting with codeSwitchingEnabled: true. However, when I configure this via the dashboard and publish the assistant: The codeSwitchingEnabled setting is not saved (it either reverts to false or disappears). The transcriber section appears disabled in the Vapi Dashboard, and I can’t manually re-enable or edit it reliably. After saving and publishing, the assistant does not behave correctly during multilingual calls. I'm using the Vapi Dashboard to configure everything — no API calls involved at this stage. Has there been a recent update affecting transcriber config or code-switching in the dashboard UI? Is there a workaround (such as using Advanced JSON or the Vapi API) to ensure codeSwitchingEnabled: true works correctly with language: "multi"? Here is the assistant I’m working with: Assistant ID: c40dd973-403e-4210-910b-bc80b6a1f491 Appreciate any help or clarification. Thanks so much! https://cdn.discordapp.com/attachments/1327398813715140669/1357690308779049181/image.png?ex=67f11f22&is=67efcda2&hm=880f270bd07afe01c1bfc2757602f0d27253bf893e7734e7bdb2bd806089b239&
By the way it was working a few hours ago. Now it just won't work! Please Help!
v
This ticket has been marked as solved, and as a result, you will no longer receive any further responses. Kindly create a new support ticket to escalate your inquiry. 
PreviousNext
Powered by