some question about a Chinese model "Yunxi".

Question

The model "zh-CN-YunxiLTSNeural" is an old model. At present, this model is not available in the document. The current model is called "zh-CN-YunxiNeural". We have tested this in Southeast Asia. Are there any other regions besides Southeast Asia? We thought the previous model sounded better. Could you help us look at it thanks

Answer 1

@FJ123 You can use the voice in different regions by looking up the following API for long audio:

https://<region-short-name>.customvoice.api.speech.microsoft.com/api/texttospeech/v3.0/longaudiosynthesis/voices  

The API for short audio is:

https://<region-short-name>.tts.speech.microsoft.com/cognitiveservices/voices/list  

For example, the voice is listed in West Europe when you run the GET query for above with the header Ocp-Apim-Subscription-Key:<key>.

https://westeurope.tts.speech.microsoft.com/cognitiveservices/voices/list  

You can try to use the voice in different region and check if it sounds consistent. Based on feedback received for voices the team updates the voice list from time to time. If the voice sounds good in one region and the other region has some static then it is most likely a technical issue with the API endpoint and this needs to be corrected by the product service team.

I hope this helps!!

If an answer is helpful, please click on or upvote which might help other community members reading this thread.