1. Installation

• Using Conda
• Using pip

Using Conda

The easiest way to install Spleeter is probably to use conda. You need first to install conda, for instance using miniconda.

⚠️ For Windows users we highly recommend to use a complete Anaconda installation. You may otherwise have to manually install ffmpeg.

We provide already provisioned packages for most platforms, you can thus install Spleeter by simply running :

conda config --add channels conda-forge # only needed if you don't already have this channel set
conda install -c deezer-research spleeter 

All dependencies required to make Spleeter work will be installed in your environment as well as the Spleeter library itself.

Note: there used to be a spleeter-gpu package that was merged into the spleeter package (following tensorflow single package for both CPU and GPU).

⚠️ For Mac users, installing Spleeter with conda will install an older version. If you want to run the latest version, you can install Spleeter with pip (see next section).

Using pip

Spleeter could also be installed directly using pip package manager, assuming that following dependencies are installed first :

• ffmpeg
• libsndfile²

² only needed if you plan to perform evaluation.

If you do not feel comfortable installing these dependencies, we recommend an installation with conda.

Note that the dependencies can be installed with conda on any platform before using pip for installing Spleeter:

conda install ffmpeg libsndfile

You can then install the library from PyPi repository :

pip install spleeter

⚠️ Running Spleeter on Nvidia GPUs devices require you to have a proper CUDA installation. We recommend to use the provided GPU Docker image.