# Audio Separator

> Audio Separator is a pip-installable CLI and Python library that runs the stem-separation models used in Ultimate Vocal Remover (MDX-Net, VR, Demucs, MDXC/RoFormer). It does two-stem vocal/instrumental splits and more stems such as drums, bass, piano and guitar, on NVIDIA CUDA, Apple Silicon, CPU or experimental DirectML, and ships Docker images. Code is MIT; licences of the non-UVR models it downloads are not stated in one place. There is no GUI, and it does not generate music.

- Repository: https://github.com/nomadkaraoke/python-audio-separator
- Official docs: https://github.com/nomadkaraoke/python-audio-separator#readme
- Checked: 2026-10-16

## What it replaces
- Suno: Splitting a song into vocal and instrument stems (Get Stems) (full). Splits any song into stems from the command line or Python; no generation.

## Facts, with sources

| Question | Status | Answer | Source | Checked |
|---|---|---|---|---|
| Runs on my machine? | Confirmed | Yes, CLI, Python package or Docker | https://github.com/nomadkaraoke/python-audio-separator#installation-%EF%B8%8F | 2026-10-15 |
| Code licence | Confirmed | MIT | https://github.com/nomadkaraoke/python-audio-separator#license- | 2026-10-15 |
| Model weights licence | Unknown | The README identifies MIT with UVR credit for UVR-trained models; a complete licence inventory for every other downloadable model, including RoFormer variants, was not verified | https://github.com/nomadkaraoke/python-audio-separator#license | 2026-10-16 |
| Commercial use | Unknown | MIT permits use of the wrapper code; commercial permission remains dependent on the selected model. The README's UVR attribution statement does not establish the terms for every downloadable weight | https://github.com/nomadkaraoke/python-audio-separator#license | 2026-10-16 |
| Paid services needed? | Confirmed | None required; an optional remote API can be self-deployed on Modal | https://github.com/nomadkaraoke/python-audio-separator#remote-api-usage- | 2026-10-15 |
| Hardware | Confirmed | CPU-only works; NVIDIA CUDA, Apple Silicon (macOS 14+) and experimental DirectML acceleration | https://github.com/nomadkaraoke/python-audio-separator#-no-hardware-acceleration-cpu-only | 2026-10-15 |
| Operating systems | Confirmed | Linux, macOS (Apple Silicon) and Windows; Docker images for amd64 and arm64 | https://github.com/nomadkaraoke/python-audio-separator#installation-%EF%B8%8F | 2026-10-15 |
| Setup effort | Confirmed | Easy; pip install audio-separator (CPU or GPU extra) or Docker; FFmpeg needed | https://github.com/nomadkaraoke/python-audio-separator#-no-hardware-acceleration-cpu-only | 2026-10-15 |

GitHub: 1420 stars, 204 forks, last push 2026-10-01, archived: no.
---
Source: RepoQuarry (https://repoquarry.dev/projects/python-audio-separator). Data licence: CC BY 4.0. Quotes are from project documentation and belong to their authors.