2.0.230618

RVC-Project/Retrieval-based-Voice-Conversion-WebUI2.0.230618Jun 5, 2024by RVC-Boss

AI Summary

This major update adds 32k and 48k pretrained models, fixes inference errors for non-f0 models, and introduces ONNX export capability. Performance optimizations for large datasets are also included.

Key Highlights

Added 32k and 48k pretrained v2 models
Fixed non-f0 model inference errors
Automatic kmeans for feature processing on datasets over 1 hour
Added ONNX export option tab
Data processing now removes outlier slices

New Features

v2 pretrained models (32k, 40k, 48k)
Vocal-to-guitar toy repository
ONNX export for inference

Full Release Notes

### 完整包 Complete package (已更新至最新版本 Has been updated to the latest version.)
https://huggingface.co/lj1995/VoiceConversionWebUI/resolve/main/RVC-beta.7z

0619update: If you're using small model extract feature, you should update [this file](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI/blob/main/train/process_ckpt.py ) because there is a small bug when config is v2-32k/48k


How to update from 0528v2 version:
1、download or clone updated codes from github, and replace 0528v2 version.
2、download new pretrained_v2 weights from https://huggingface.co/lj1995/VoiceConversionWebUI/tree/main/pretrained_v2 (32k and 48k weights needed. 40k has already been supported in 0528v2 version)

Changelog:

- v2增加32k和48k两个新预训练模型
- 修复非f0模型推理报错
- 对于超过一小时的训练集的索引建立环节，自动kmeans缩小特征处理以加速索引训练、加入和查询
- 附送一个人声转吉他玩具仓库
- 数据处理剔除异常值切片
- onnx导出选项卡

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