Skip to content

Latest commit

 

History

History
137 lines (105 loc) · 8.96 KB

README_zh.md

File metadata and controls

137 lines (105 loc) · 8.96 KB
logo

MSST-WebUI

English | 简体中文 | 繁體中文 | 日本語 | 한국어
一个为Music-Source-Separation-Training打造的WebUI应用,此外我们也添加了UVR进来!
Open in Google Colab GitHub release GitHub stars GitHub license Hugging Face Model

介绍

这是一个用于 Music-Source-Separation-Training (MSST) 的WebUI,MSST是一个用于训练音乐源分离模型的仓库。您可以使用这个WebUI推断MSST模型和VR模型,预设流程页面允许您自定义处理流程。您可以在“安装模型”界面安装模型。如果您之前已经下载了 Ultimate Vocal Remover (UVR),则无需重新下载VR模型。您可以直接进入“设置”页面,选择您的UVR5模型文件夹。我们还在WebUI中提供了一些便捷的工具,例如 Singing-Oriented MIDI Extractor (SOME)、更高级的合奏模式等。

使用

  • Windows: 从 Releases 下载安装程序,或克隆此仓库并从源代码运行。
  • Linux/macOS: 克隆此仓库并从源代码运行。
  • Google Colab: 点击这里 在Google Colab上运行WebUI。

下载链接

推荐国内用户使用后三个下载链接,无需魔法即可下载。此外,我们向国内用户提供详细的飞书教程文档

网站 下载链接 提取码 备注
Github Releases https://github.com/SUC-DriverOld/MSST-WebUI/releases - 仅提供安装程序,无模型下载
Huggingface https://huggingface.co/Sucial/MSST-WebUI/tree/main - 安装程序及所有可用模型
Huggingface镜像站 https://hf-mirror.com/Sucial/MSST-WebUI/tree/main - 安装程序及所有可用模型
百度网盘 https://pan.baidu.com/s/1uzYHSpMJ1nZVjRpIXIFF_Q 1145 安装程序及所有可用模型
123网盘 https://www.123pan.cn/s/1bmETd-AefWh.html 1145 安装程序及所有可用模型

文档

我们为中文用户提供了一些详细的中文文档,点击 这里 跳转。对于其他语言的用户,前往 docs 文件夹查看部分文档。你也可以参考 deton24Instrumental and vocal & stems separation & mastering guide

从源码运行

  • 克隆仓库

    git clone https://github.com/SUC-DriverOld/MSST-WebUI.git
    cd MSST-WebUI
  • 使用Conda创建Python虚拟环境,此处建议使用Python 3.10。

    conda create -n msst python=3.10 -y
    conda activate msst
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
    pip install -r requirements.txt --only-binary=samplerate
  • 安装完依赖后,进入 site-packages 文件夹,打开 librosa\util\utils.py 文件,定位到第 2185 行。将该行 np.dtype(complex): np.dtype(np.float).type, 修改为 np.dtype(complex): np.dtype(float).type,。如果您不知道如何操作,可以使用以下命令。

    pip uninstall librosa -y
    pip install tools/webUI_for_clouds/librosa-0.9.2-py3-none-any.whl
  • 使用下面的命令启动WebUI。

    python webUI.py
  • 如需更详细的启动参数,或如果您想在云平台上运行WebUI,请查看 此文档(英文)

  • 使用模型类型 swin_upernet 时,您可能会遇到以下错误:ValueError: Make sure that the channel dimension of the pixel values match with the one set in the configuration。请参考 此issue 解决。

CLI & Python API

参考 此文档(英文) 获取更多详细信息。

训练 & 验证

参考 此文档(英文) 获取更多详细信息。

参考

感谢所有贡献者的共同努力

感谢所有的模型提供者

anvuew aufr33 deton24 jarredou pcunwa Sucial
anvuew aufr33 deton24 jarredou pcunwa Sucial
Super-YH viperx wesleyr36 yxlllc ZFturbo
Super-YH viperx wesleyr36 yxlllc ZFturbo