ga642381 / Taiwanese-Speech-Synthesis

Taiwanese Speech Synthesis with Tacotron2

taiwanese-speech-synthesis

• This is a Taiwanese Speech Synthesis System based on Tacotron2 and WaveRNN
• The input is in 臺羅拼音，and the output is Taiwanese Speech
• The model that translate from Chinese(中文，華語) to 臺羅拼音 will be released soon
• The trained TTS model can be combined with TaiwaneseTTS, so that we will have a awesome GUI interface. I intend to merge these two repositories into one in the future

Quick Start

1. Download pretrained models
2. Specify the inputs in "sentences.txt"
3. Generate audios with the command

python gen_tts.py sentences.txt --tts_weights ./pretrained_models/tacotron2.pyt --voc_weights ./pretrained_models/wavernn.pyt --save_dir ./results

References

1. 臺灣言語工具
2. fatchord/WaveRNN
3. i3thuan5/SuiSiann-HunLian
4. 台灣媠聲(SuíSiann Dataset)