Open furqan4545 opened 11 months ago
This is probably because this VALLE-E-X model wasn't trained on the same amount of data, not for as long. Hopefully someone trains a model on the full librilight dataset soon
I've managed to create a finetuning colab on my fork... hopefully ill get around to training
@korakoe did you ever manage to get good quality out of this?
I tried to clone many voices but it failed all the time. Was just spitting out a useless cloned voice and sometime not even speaking properly.