Starstruk Logo MarkStarstruk

1.2.230428

RVC-Project/Retrieval-based-Voice-Conversion-WebUI1.2.230428Jun 5, 2024by RVC-Boss

AI Summary

This release improves faiss index performance and removes GPU limitations. Bug fixes for UVR5 and expanded mini-GUI support are included.

Key Highlights

  • Upgraded faiss index settings - faster and higher quality
  • Removed total_npy dependency - easier model sharing
  • Unlocked 16-series GPU limits - 4GB VRAM GPUs get full 4GB settings
  • Fixed UVR5 vocal/instrument separation bugs
  • Real-time voice changing mini-GUI now supports non-40k models

New Features

  • Improved faiss index performance
  • Support for 4GB VRAM GPUs
  • Mini-GUI support for various sample rates and pitch models

Full Release Notes

### 完整包 base 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

### 20230428 更新包 updated package
> 下过老完整包的用户可以只下更新包 Users that downloaded old base package can use the updated package.

见下方`Assets`, 解压到RVC根目录覆盖完整包下的一些文件 Unzip it in RVC root and replace some files of old version.

### 更新日志

### 20230428更新
- 升级faiss索引设置，速度更快，质量更高
- 取消total_npy依赖，后续分享模型不再需要填写total_npy
- 解锁16系限制。4G显存GPU给到4G的推理设置。
- 修复部分音频格式下UVR5人声伴奏分离的bug
- 实时变声迷你gui增加对非40k与不懈怠音高模型的支持


### 后续计划：
功能：
- 增加选项:每次epoch保存的小模型均进行提取
- 增加选项:推理额外导出mp3至填写的路径
- 支持多人训练选项卡（至多4人）

底模：
- 收集呼吸wav加入训练集修正呼吸变声电音的问题
- 我们正在训练增加了歌声训练集的底模，未来会公开
- 升级鉴别器
- 升级自监督特征结构