Hi, congrats on your work!
May I ask about the training dataset, which datasets do you use for training?
And which datasets do you use for training speaker encoder, does the speaker encoder is also trained through over 100k hours.