A tensofrlow implementation of 'Neural TTS Stylization with Adversarial and Collaborative Games', ICLR 2019
Python >= 3.0
Tensorflow == 1.4.0
Other libraries can be installed by running:
pip3 install -r requiredments.txt
You can use our pre-trained model to test any text you want.
You can specify any styles (emotion) by replacing the reference audio, and input any text you want by adding your text in a script '*.txt'.
We offer some examples for you test in a quick start.
1.The example texts are in 'examples_test.txt'
2.The example reference audios in './reference_audios'
3.The pre-trained model is in './pretrained-model'
To synthsizing, you need to run:
python3 test.py --model pretrained-model/model.ckpt-161000 --reference reference_audios/*.wav
To synthsizing by your own model:
python3 test.py --model 'YOUR_MODEL_PATH' --reference 'YOUR_REFERENCE_AUDIO_PATH'
#1. Preprocess data
Write your training list in this format:
'mel-spec-path'|'linear-spec-path'|'lenth'|'text'|'label'
Run:
python3 train.py
We set the default hyperparameters the same with Google's paper ' Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis'.
You can adjust your own hyparameters in hparams.py
Reference:
Shan Yang's implementation of GST-Tacotron: https://github.com/syang1993/gst-tacotron
Yuxuan Wang, Daisy Stanton, Yu Zhang, RJ Skerry-Ryan, Eric Battenberg, Joel Shor, Ying Xiao, Fei Ren, Ye Jia, Rif A. Saurous. 2018. Style Tokens: Unsupervised Style Modeling, Control and Transfer in End-to-End Speech Synthesis
RJ Skerry-Ryan, Eric Battenberg, Ying Xiao, Yuxuan Wang, Daisy Stanton, Joel Shor, Ron J. Weiss, Rob Clark, Rif A. Saurous. 2018. Towards End-to-End Prosody Transfer for Expressive Speech Synthesis with Tacotron.