The training data used for WavTokenizer-small

[dzr1026]

I would like to ask whether the dataset used to train WavTokenizer-small is the complete LibriTTS or the LibriTTS-580h mentioned in the paper. I am planning to reproduce this work and would like to know how much data is needed to achieve the current performance.

[dzr1026]

Thank you for your reply. Could you explain how LibriTTS-580h is composed? Is it formed by combining train-clean-100h, train-clean-360h, and parts of train-other-500h?

[jishengpeng]

Thank you for your reply. Could you explain how LibriTTS-580h is composed? Is it formed by combining train-clean-100h, train-clean-360h, and parts of train-other-500h?

Thanks for your attention. LibriTTS-585hours (all train-clean-100 version+train-clean-360 version+train-other-500 version, LibriTTS is subset of LibriSpeech and total hours are 585. It is very easy to reproduce the small-version.

[dzr1026]

Thank you !!