Spleeter

What is Spleeter?

Deezer, a music streaming platform, has open-sourced Spleeter, an audio separation library built on Python and TensorFlow that uses machine learning to efficiently isolate vocals from any song. Spleeter is a powerful tool that enables users to divide a music track into separate components, such as vocals, drums, bass, and other specific sounds. The library provides pretrained state-of-the-art models for performing various flavors of separation, including vocals/accompaniment (2 stems), vocals/drums/bass/other (4 stems), and vocals/drums/bass/piano/other (5 stems). The 2-stem and 4-stem models have demonstrated impressive performance on the musdb dataset

Highlights

• Ability to separate audio tracks into individual components (vocals, drums, bass, etc.)
• Provides pre-trained models for different separation configurations (2 stems, 4 stems, 5 stems)
• Supports high-speed separation, up to 100x faster than real-time when run on a GPU
• Can be used via command line or integrated into development pipelines as a Python library
• Easy to install and use, with support for Conda, pip, and Docker