Great work! I the can't get the right demo.pkl by make_metadata.py！

auspicious3000 / SpeechSplit

Unsupervised Speech Decomposition Via Triple Information Bottleneck

http://arxiv.org/abs/2004.11284

MIT License

644 stars 92 forks source link

Closed c1a1o1 closed 4 years ago

c1a1o1 commented 4 years ago

Great work! The question is to get Mel, normed-F0 value!

Who can help me? [Speaker_Name , One-hot , [Mel, normed-F0, length, utterance_name] ]

auspicious3000 commented 4 years ago

The code is self-explanatory. Please make sure to carefully read the code related to demo.pkl