Custom Neural Voice Training Problem III

Hero Li(李柏緯) 40

Hello, everyone,

I have a question regarding the Custom Neural Voice (CNV) training process.

Our goal is to train a model to speak both Mandarin and English. Our dataset primarily consists of Taiwanese individuals who speak Mandarin Chinese (Traditional) as their primary language, with a smaller amount of English data.

Here's my question:

If I choose the Neural Cross-Linguistic model, will the model be capable of speaking both the languages used during training and the target language, or will it only be able to speak the target language alone?

Your insights on this matter would be greatly appreciated.

Thank you for your time.

romungi-MSFT 48,401 Reputation points Microsoft Employee

2023-10-27T10:40:10.5966667+00:00

@Hero Li(李柏緯) I think I have answered this query in one of your previous question. If you choose traditional chinese you can use a secondary language that also supports bi-lingual and use it for speaking english. So, with the zh-CN data, you can create a voice that speaks en-AU or any of the languages with Cross-lingual support.
Hero Li(李柏緯) 40 Reputation points

2023-11-21T03:04:42.9266667+00:00

We chose the latest model (type = V9.2023.10) with training data = Chinese (Taiwanese Mandarin). It can speak both Mandarin and English now.
Hero Li(李柏緯) 40 Reputation points

2023-11-21T03:05:47.3466667+00:00

We chose the latest model (type = V9.2023.10) with training data = Chinese (Taiwanese Mandarin). It can speak both Mandarin and English now.

Share via

Custom Neural Voice Training Problem III

Your answer