this work original implement by kyubyong, I changed it into my version. Thanks for the original author. And of course this work is based on Google Original Paper.
- Python3
- Tensorflow 1.2 (the newest version)
- Python Package: librosa, tqdm
To start to train tacotron, should download audio file from here, and the accordingly text file from here. These two file can be changed into your own training data. For other language, only prepared the audio data and the text data, the model can be trained without even change the python codes.
To using model to convert text to audio, just run the eval.py
, and give your text, you will get some examples of audios.
If you want to train an TTS model for Chinese, you can add me wechat, we are working on that: jintianiloveu
.
This Code OpenSource Under Apache 2.0 License.