Training a mixed Chinese and English timbre model

[pipijiev12]

Currently, I use the example/libritts/cosyvoice/run.sh script to train the model. I found that libritts is an all-English dataset. If I need to train a new mixed Chinese and English voice model, can I just add the Chinese and English datasets of the corresponding characters to libritts for training?

[aluminumbox]

yes you can, you need mixed CH/EN data, not just pure CH and pure EN data

[pipijiev12]

yes you can, you need mixed CH/EN data, not just pure CH and pure EN data

Does the dataset format require special handling?