Starstruk Logo MarkStarstruk

2.1.230814

RVC-Project/Retrieval-based-Voice-Conversion-WebUI2.1.230814Jun 5, 2024by RVC-Boss

AI Summary

This release introduces the RMVPE pitch extraction model - the strongest open-source vocal pitch extraction model - and adds AMD/Intel GPU support through pytorch-dml. Multiple bug fixes and quality-of-life improvements are included.

Key Highlights

  • Introduced RMVPE - strongest open-source vocal pitch extraction model
  • AMD/Intel GPU support via pytorch-dml for real-time inference and separation
  • Fixed paths with spaces support for training
  • Fixed CPU issue with index in real-time voice changing
  • Changed faiss save path from absolute to relative paths

New Features

  • RMVPE supports pytorch/onnx/DirectML
  • Real-time voice conversion on AMD/Intel GPUs
  • Vocal/instrument separation with GPU memory cleanup
  • Training now switches to CPU when GPU not supported

Full Release Notes

### 完整包 Complete package

For Nvidia GPU users:
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC0813Nvidia.7z

For AMD/Intel GPU users:
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC0813AMD_Intel.7z

中国用户可使用以下2个源加速下载
1、无需登录，免费满速下载链接https://www.123pan.com/s/5tIqVv-QHNcv.html
2、有度盘超级会员的可以用度盘https://pan.baidu.com/s/19530AOh2H3Feuti_D51cXw?pwd=reqy

Changelog [(English verison)](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/docs/en/Changelog_EN.md):

1-常规bug修复
 - 保存频率总轮数最低改为1 总轮数最低改为2
 - 修复无pretrain模型训练报错
 - 增加伴奏人声分离完毕清理显存
 - faiss保存路径绝对路径改为相对路径
 - 支持路径包含空格（训练集路径+实验名称均支持，不再会报错）
 - filelist取消强制utf8编码
 - 解决实时变声中开启索引导致的CPU极大占用问题

2-重点更新
 - 训练出当前最强开源人声音高提取模型RMVPE，并用于RVC的训练、离线/实时推理，支持pytorch/onnx/DirectML
 - 通过pytorch-dml支持A卡和I卡的
（1）实时变声（2）推理（3）人声伴奏分离（4）训练暂未支持，会切换至CPU训练；通过onnx_dml支持rmvpe_gpu的推理

Please look forward to the pretrained base model of RVCv3, which has larger parameters, larger training data, better results, unchanged inference speed, and requires less training data for training.