Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

StemDeck

github.com

  • Free plan
  • Windows
  • macOS
  • Linux
  • Web
  • API
  • Self-hosted
7.6 Club score Visit github.com

What StemDeck is

StemDeck separates songs into stems for practice, study, transcription, remixing, and other creative audio work. Its Demucs htdemucs_6s model separates vocals, drums, bass, guitar, piano, and other sounds. Users can import MP3, WAV, FLAC, OGG/Opus, MP4, and M4A files or use YouTube URLs. The workspace provides waveform lanes, mute and solo controls, volume adjustment, and looping, with exports for individual stems or a custom mix. Song analysis includes BPM, key, scale, confidence, integrated LUFS, and sample peak level. StemDeck says audio processing runs locally and audio is not transmitted, stored, cached, or redistributed by the app. Internet access is needed for YouTube downloads and the initial model fetch of about 170 MB; the model is cached afterward. Deployment options include Docker, Unraid, and running the web server from source, and the local server exposes REST endpoints. The free Apache-2.0-licensed project has no account, quota, or subscription, but handles one job at a time and up to six stems. It has no mobile app, and processing speed depends on local hardware.

Who it suits

StemDeck may suit musicians, producers, teachers, hobbyists, and listeners who want to practice, study, transcribe, or remix audio. It is intended for people comfortable working with a local tool or self-hosted setup.

Good to know, and what to watch

  • Separates six kinds of stems, including vocals and instruments
  • Accepts common audio and video formats plus YouTube URLs
  • Exports individual stems or a custom mix
  • Analyzes BPM, key, scale, and audio levels
  • Processes audio locally, according to the project
  • Handles one job at a time
  • Supports up to six stems
  • No mobile app
  • Initial model fetch requires about 170 MB of internet data

StemDeck

Free

Free plan

The verdict

StemDeck offers local stem separation alongside mixing controls and song analysis, with a free plan and no account requirement. Its one-job limit and hardware-dependent processing speed are worth considering for larger workloads.

What the maker publishes

API
The local server exposes REST endpoints for jobs, status, cancellation, stem streaming, and video export, plus server-sent job events. source
Audience
StemDeck is aimed at musicians, producers, teachers, hobbyists, and listeners who want a focused tool for practice, study, or remixing. source
Audio analysis
Song analysis includes BPM, key, scale, confidence, integrated LUFS, and sample peak level. source
Import formats
It accepts MP3, WAV, FLAC, OGG/Opus, MP4, and M4A files, as well as YouTube URLs. source
Internet needs
Internet is needed for YouTube downloads and the initial model fetch of about 170 MB; the model is cached afterward. source
Mixer and editing
The workspace includes waveform lanes, mute and solo controls, volume adjustment, looping, and exports of individual stems or a custom mix. source
Notable limits
The project describes one job at a time, up to six stems, and no mobile app; it also says processing speed depends on local hardware. source
Platforms
Prebuilt desktop downloads are offered for macOS and Windows, and the web server can be run on macOS, Linux, or Windows. source
Privacy
StemDeck says audio processing runs locally and audio is not transmitted, stored, cached, or redistributed by the app. source
Purpose
StemDeck separates songs into stems for practice, study, transcription, remixing, and creative audio workflows. source
Self-hosting
The project documents Docker and Unraid deployment as well as running the web server from source. source
Stem types
The Demucs htdemucs_6s model separates vocals, drums, bass, guitar, piano, and other sounds. source
Support and project
The maker links to GitHub issues and pull requests, Discord, Reddit, Instagram, and X; the project is licensed under Apache-2.0. source

Checked against the maker's pages on 30 Sept 2026.