ASR training data must be same with stargan training data?

[joan126]

Hi, I am trying to training starganv2 vc , one question about dataset for training ASR and vc, can I use different dataset for ASR training and starggan ?

[yl4579]

It can be totally different, even in different languages. The released pre-trained model was trained on LibriTTS and the voice conversion model was trained on VCTK. The same ASR model works for JVS datset (Japanese) too.

[joan126]

It can be totally different, even in different languages. The released pre-trained model was trained on LibriTTS and the voice conversion model was trained on VCTK. The same ASR model works for JVS datset (Japanese) too.

I am working on Mandarin Chinese. will the pretrained ASR model works for this?

[yl4579]

I haven't tested on Mandarin yet, but I think it should, though the quality would not be as good as if you train it directly with Mandarin. You can refer to this issue for ASR training and data preparation in Mandarin: yl4579/AuxiliaryASR#2

[joan126]

I haven't tested on Mandarin yet, but I think it should, though the quality would not be as good as if you train it directly with Mandarin. You can refer to this issue for ASR training and data preparation in Mandarin: yl4579/AuxiliaryASR#2

thanks a lot!!