Spleeter
An open-source audio separation library that extracts vocals and instruments locally with pretrained TensorFlow models and supports Docker and GPU processing.
Spleeter is Deezer's music separation library for developers and audio researchers who want to extract individual parts from a recording on their own hardware. Its pretrained models let you process music without training a model first.
- Vocal isolation: Separate a singing voice from its accompaniment to produce two audio tracks.
- Instrument separation: Extract vocals, drums, bass and other sounds, with a five-track model that also separates piano.
- Application integration: Use the command-line tool for audio files or the Python library within an audio processing pipeline.
- Custom training: Train your own separation models when you have a dataset of isolated sound sources.
Spleeter uses Python and TensorFlow, supports Docker and can process audio on a GPU. Deezer reports four-track separation at up to 100 times real-time speed on a GPU; that figure describes GPU processing rather than a speed guarantee for every machine.
The project is open source under the MIT license. Its pretrained models have been used in iZotope RX, SpectralLayers, Acoustica and VirtualDJ, and Spleeter 4 Max brings the technology into Ableton Live. Deezer also offers a separate commercial product, Spleeter Pro, with professional support. Third-party websites and graphical apps built around Spleeter aren't hosted, maintained or directly supported by Deezer.
Logotyper
