- Python 3.6
- Numpy
- Scipy
- Tensorflow >= 1.4.1
- librosa
- pysptk
- soundfile
- matplotlib
- wavenet_vocoder:
pip install wavenet_vocoder == 0.1.1
Download training data from the CSTR VCTK corpus to assets.
- Extract spectrogram and f0:
python make_spect_f0.py - Generate training metadata:
python make_metadata.py - Run training scripts:
python main.py