Ultimate Vocal Remover (UVR)

Resources

GitHub

Official documentation

Bugs (upstream)

Ultimate Vocal Remover (UVR) is software that separates vocals from instrumentals in music audio files using deep neural networks and separation models.

Installation

Emerge

At this time, there is no known ebuild for this software. It can be installed manually as follows.

Manual installation

Reference: https://github.com/Anjok07/ultimatevocalremovergui/?tab=readme-ov-file#linux-installation

Portage packages

The dev-lang/python package needs to have the tk USE flag.

FILE /etc/portage/package.use/package.usepackage.use example

dev-lang/python tk

If it does not, re-emerge it:

root #emerge --ask --oneshot dev-lang/python

If there are multiple versions of dev-lang/python installed, be sure that the Python version that will be re-emerged corresponds to the current Python version.

user $python --version

If it isn't, specify the version:

root #emerge --ask --oneshot =dev-lang/python-3.11.8_p1

Get source code

Obtain the UVR source code. Either with dev-vcs/git, or downloading a zip archive. Do this as a non-root user.

With git:

user $git clone https://github.com/Anjok07/ultimatevocalremovergui.git

or download an archive:

user $wget https://github.com/Anjok07/ultimatevocalremovergui/archive/refs/heads/master.zip

and unpack the zip file (emerge app-arch/zip if necessary).

Build from source

Enter the source directory, which is either the path to the ultimatevocalremovergui repository that was cloned, or the directory where the zip file was unzipped.

user $cd ultimatevocalremovergui

Modify the requirements.txt file to remove an unused dependency which causes installation problems:

user $${EDITOR} requirements.txt

FILE ./requirements.txtPython dependency modifications

# remove this line
Dora==0.0.3

It may be necessary to modify numpy to a higher version number:

FILE ./requirements.txtPython dependency modifications

# change this line
numpy==1.23.5
# to a slightly higher version, such as
numpy==1.25.2

Create a Python virtual environment:

user $python -m venv venv

Activate the virtual environment:

user $. venv/bin/activate

Update the wheel package (safe to do if no upgrade is needed):

user $pip install --upgrade wheel

Make the Python dependency installation script runnable:

user $chmod +x install_packages.sh

and run it:

user $./install_packages.sh

Make sure there are no errors (and not just at the end of the output).

Usage

Invocation

user $python UVR.py

To exit the Python virtual environment when done:

user $deactivate

Configuration

Configuration is done within the GUI. There are many ways to use UVR, but here is one way to configure it to do basic processing.

• Choose an input audio file.
• Choose an output directory.
• Under "CHOOSE PROCESS METHOD", select MDX-Net.
• Leave SEGMENT SIZE and OVERLAP as is (256, and Default).
• Under CHOOSE MDX-NET MODEL, select Download More Models. A settings window will appear.
• Download the "UVR-MDX-NET Inst Main" model.
• Close the settings window.
• Ensure the "UVR-MDX-NET Inst Main" model is selected in the dropdown.

Click the "Start Processing" button.