Skip to content

Singing Voice Conversion via diffusion model

License

Notifications You must be signed in to change notification settings

Jnstiu/diff-svc

Repository files navigation

Diff-SVC

Singing Voice Conversion via diffusion model

update:

现已支持对自定义数据集的训练
修改config中的配置,预处理时将自动将统计数据写入config中,若需要移动config位置,请修改config中对自身位置的引用
wav数据直接放在raw_data_dir下即可,自适应任何目录结构 其他配置项如有需要请参考config中的名称修改

推理:

查看./inference.ipynb

预处理:

export PYTHONPATH=.
CUDA_VISIBLE_DEVICES=0 python preprocessing/binarize.py --config training/config.yaml

训练:

CUDA_VISIBLE_DEVICES=0 python run.py --config training/config.yaml --exp_name [your project name] --reset

已训练模型

checkpoints trained on opencpop dataset(and others in future) can be found here(QQ channel)
please scan the QRcode with QQ:

Acknowledgements

项目基于diffsinger原仓库diffsinger(openvpi维护版)开发.
十分感谢openvpi成员在开发训练过程中给予的帮助。

About

Singing Voice Conversion via diffusion model

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 83.3%
  • Python 16.7%