How to select audio language?
Hi,
How to make two wav files, one in Cantonese audio (Yue Chinese), one in Chinese audio (Chinese [Mandarin]).
For example: Chinese Text: "人工智能超速發展"
Any parameter for audio select?
Thanks
Hi @SkyNetWalker ,
If I understand correctly, you’re encountering pronunciation issues when the reference audio and the intended output are in different languages—for example, when using a Cantonese reference to generate Mandarin speech, or vice versa.
This is a known limitation. Cantonese and Mandarin share many written characters, and our training data contains substantially more Mandarin than Cantonese. As a result, the model may sometimes favor Mandarin pronunciation in these cases.
For the best results, we recommend using reference audio in the language you want to generate. For your example, “人工智能超速發展”, use a Cantonese reference to generate the Cantonese version and a Mandarin reference to generate the Mandarin version. The reference audio provides contextual cues that help guide the model toward the intended pronunciation.
Hi @SkyNetWalker ,
If I understand correctly, you’re encountering pronunciation issues when the reference audio and the intended output are in different languages—for example, when using a Cantonese reference to generate Mandarin speech, or vice versa.
This is a known limitation. Cantonese and Mandarin share many written characters, and our training data contains substantially more Mandarin than Cantonese. As a result, the model may sometimes favor Mandarin pronunciation in these cases.
For the best results, we recommend using reference audio in the language you want to generate. For your example, “人工智能超速發展”, use a Cantonese reference to generate the Cantonese version and a Mandarin reference to generate the Mandarin version. The reference audio provides contextual cues that help guide the model toward the intended pronunciation.
hi @chua
Thank you for your suggestion, i will try it later. 😉