Skip to content

KevinWang676/VITS2-Chinese

Repository files navigation

VITS2 Chinese 🎶🌟💕 (Work in Progress)

只需上传一段语音素材,程序自动将语音切片、标注、预处理,一键训练

环境配置

  1. 运行
git clone https://github.com/KevinWang676/VITS2-Chinese.git
cd VITS2-Chinese
pip install -r requirements.txt
  1. 运行
cd monotonic_align
python setup.py build_ext --inplace
  1. 上传语音文件:请上传一段中文单说话人的语音文件,建议为较大数据集对应的.wav文件

语音处理

  1. 语音切片:在filename处填写上传的语音文件名
python split.py --filename {filename}.wav
  1. 语音标注:标注完成后,可以在filelists/short_character_anno.list文件中对标注内容微调
python short_audio_transcribe.py --languages "C" --whisper_size large
  1. 语音预处理
python preprocess.py

训练及推理

  1. 开始训练
python train.py -c ./configs/config.json -m OUTPUT_MODEL
  1. 推理

参考inference.ipynb

Acknowledgement

Based on MB-iSTFT-VITS2

About

VITS2 for Chinese speech | 最新VITS2中文语音合成

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published