NOTE: This is just a clone of r9y9's awesome repository. We have just added a few variations to train.py to accomodate our experiments. These versions each work for a different approach in our project. The usage is described here. Please note that the train scripts do not change the original setup instructions or training commands in any manner. They are added just to aid in reproducibility of our experiments.
PyTorch implementation of Tacotron speech synthesis model.
Inspired from keithito/tacotron. Currently not as much good speech quality as keithito/tacotron can generate, but it seems to be basically working. You can find some generated speech examples trained on LJ Speech Dataset at here.
If you are comfortable working with TensorFlow, I'd recommend you to try https://github.com/keithito/tacotron instead. The reason to rewrite it in PyTorch is that it's easier to debug and extend (multi-speaker architecture, etc) at least to me.
- PyTorch
- TensorFlow (if you want to run the training script. This definitely can be optional, but for now required.)
git clone --recursive https://github.com/r9y9/tacotron_pytorch
pip install -e . # or python setup.py develop
If you want to run the training script, then you need to install additional dependencies.
pip install -e ".[train]"
The package relis on keithito/tacotron for text processing, audio preprocessing and audio reconstruction (added as a submodule). Please follows the quick start section at https://github.com/keithito/tacotron and prepare your dataset accordingly.
If you have your data prepared, assuming your data is in "~/tacotron/training" (which is the default), then you can train your model by:
python train.py
Alignment, predicted spectrogram, target spectrogram, predicted waveform and checkpoint (model and optimizer states) are saved per 1000 global step in checkpoints directory. Training progress can be monitored by:
tensorboard --logdir=log
Open the notebook in notebooks directory and change checkpoint_path to your model.