Spleeter: AI-Powered Music Source Separation by Deezer
Learn how Spleeter by Deezer uses TensorFlow to isolate vocals, drums, bass, and more from any audio track. Open source, MIT license, Docker-ready.
What Is Spleeter?
Spleeter is an open-source source separation library developed by Deezer that uses pretrained TensorFlow models to split audio tracks into individual components. Whether you are a music producer isolating vocals for a remix, a podcaster cleaning up audio, or a researcher studying musical structure, Spleeter delivers professional-grade separation with a single command. With 28k+ GitHub stars, it has become the industry standard for AI-powered music demixing, used by iZotope RX 8, Steinberg SpectralLayers, VirtualDJ, and Algoriddim djay.
Key Features
Spleeter supports three levels of separation: 2 stems (vocals vs accompaniment), 4 stems (vocals, drums, bass, other), and 5 stems (adding piano). On a GPU it processes audio 100x faster than real-time — a 3-minute song separates into 4 stems in under 2 seconds. The library is Python-based on TensorFlow and integrates easily into any audio pipeline.
Why Self-Host Spleeter?
Running Spleeter yourself means no files leave your machine, no API limits, no usage fees. Process unlimited tracks, batch entire libraries, and customize parameters. The Docker image makes deployment trivial on any Linux server or workstation with a GPU.
Use Cases
- Music Production: Isolate vocals for remixes, create instrumentals
- Podcasting: Clean up background music, enhance voice clarity
- Content Creation: Extract stems for video editing, make karaoke tracks
- Research: Music information retrieval and audio processing
Conclusion
Spleeter puts professional-grade source separation into the hands of anyone with Docker. Its speed, accuracy, and open-source MIT license make it an essential AI audio tool.