Demucs

By Meta AI research ยท Stem separation

About Demucs

Open-source separation model from Meta research that splits music into drums, bass, vocals and other.

Demucs is an open-source music source separation model published by Meta's AI research team. Given a mixed song, it separates drums, bass, vocals and everything else into their own files, and some of its models also split out guitar and piano. Many friendlier apps use Demucs or similar models under the hood.

You run it yourself with Python from the command line, on a computer with enough memory, and a graphics card makes it much faster. That makes it a good fit for a band member who is comfortable with a terminal and wants to process a batch of rehearsal or archive recordings without uploading them anywhere. If you would rather click a button, a hosted tool or the Ultimate Vocal Remover desktop app gives you similar models with a graphical interface. Check the license of the model files before building it into a commercial product.

Details

  • Free plan or free version

    Yes
  • Open source

    Yes

Not enough info yet on works in the browser, desktop app or plugin, mobile app. Check the tool's own site.

Listing written by PerfectBands, checked 2026-11-03. Report a correction or request removal. Back to stem separation.