README.md

Speech Synthesis

Currently supported TTS tasks are LJSpeech and Chinese Standard Mandarin Speech Copus(data_baker). Supported models are shown in the table below: (Note:HiFiGAN is trained based on TensorflowTTS)

Traing Data	Acoustic Model	Vocoder	Audio Demo
data_baker	Tacotron2	GL	audio_demo
data_baker	Transformer_tts	GL	audio_demo
data_baker	Fastspeech	GL	audio_demo
data_baker	Fastspeech2	GL	audio_demo
data_baker	Fastspeech2	HiFiGAN	audio_demo
ljspeech	Tacotron2	GL	audio_demo

To perform the full procedure of TTS experiments, simply run:

source tools/env.sh
bash examples/tts/$task_name/run.sh

Core Stages:

Take Tacotron2 as an example.

####Note: You should install unrar at first.

apt-get update
apt-get install unrar

(1) Data preparation

You can run examples/tts/$task_name/run.sh, it will download the corresponding dataset and store it in examples/tts/$task_name/data. The script examples/tts/$task_name/local/prepare_data.py would generate the desired csv file decripting the dataset

(2) Data normalization

With the generated csv file, we should compute the cmvn file firstly like this

$ python athena/cmvn_main.py examples/tts/$task_name/configs/t2.json examples/tts/$task_name/data/train.csv

We can also directly run the training command like this

$ python athena/main.py examples/tts/$task_name/configs/t2.json

It will automatically compute the cmvn file before training if the cmvn file does not exist.

(3) Training

You can train a Tacotron2 model using json file examples/tts/$task_name/configs/t2.json

(4) Synthesis

Currently, we provide a simple synthesis process using GriffinLim vocoder. To test your training performance, run

$ python athena/inference.py examples/tts/$task_name/t2.json

(5) SavedModel

You can also save your ckpt as a TensorFlow SavedModel:

$ python athena/freeze.py examples/tts/$task_name/t2.json your_saved_model_dir