This repository is forked from xcmyz's implementation and updates the model by incorporating a neural discriminator to implement an Alpha GAN architecture.
- Download and unzip LJSpeech dataset.
- Put LJSpeech dataset under
data
folder. - Unzip
alignments.zip
. - Put Nvidia pretrained waveglow model under
waveglow/pretrained_model
folder. - Download the trained Alpha GAN model and put it under
model_alpha_gan
folder. - Download the trained FastSpeech model and put it under
model_fastspeech
folder. - Run
python3 preprocess.py
.
Run python3 train.py
.
- Run
python3 eval.py
.
- Change the
checkpoint_path
inhparams.py
to eithermodel_alpha_gan
ormodel_fastspeech
. - Run
python3 eval.py
.
Run python3 plot.py