Spleeter

Spleeter: Open-Source Audio Source Separation Tool

Spleeter separates audio recordings into stems of vocals, drums, bass, and other instruments for remixing or analysis, utilizing deep learning for high quality source separation.

What is Spleeter?

Spleeter is an open-source audio source separation Python library created by Deezer Research for separating audio recordings into stems of vocals, drums, bass, piano and other instruments. It uses state-of-the-art machine learning techniques such as convolutional neural networks and trained models for performing music source separation efficiently in the browser. The pre-trained models automatically estimate the origin stems from an audio file, dramatically speeding up the editing process.

Some key features and capabilities of Spleeter include:

• Separates vocals, drums, bass and other instruments from stereo music recordings into individual waveform files
• Pre-trained models using TensorFlow for fast and high-quality source separation in 5-stems or 2-stems
• Command line interface and Python API for integration into production workflows
• Custom separation model training on new datasets
• Active open source development by Deezer Research and contributions from a community of musicians and audio engineers

Spleeter is a useful open source tool for tasks like music manipulation, remixing, upmixing or analysis. For example, it can be used by musicians to quickly isolate the vocals or a specific instrument track from a song for sampling/remixing, or by audio engineers for tasks like music production, audio post-production and audio restoration. As an ML-based tool it keeps improving, but may not be on par yet with dedicated commercial solutions.

Spleeter Features

Official Links