StemDeck, a free, open-source and local AI stem separator

StemDeck is a free, open-source stem separation tool that allows users to separate audio files into up to six stems: vocals, drums, bass, guitar, piano, and other. It can process various audio file formats, such as MP3, WAV, FLAC, OGG/Opus, MP4, and M4A, as well as YouTube URLs. Users can upload an audio file directly onto the import bar or paste a YouTube URL, and StemDeck will split the audio into stems.

The tool can be used in a DAW-style multitrack mixer, where users can mute, solo, balance levels, zoom, loop, and export individual stems or a custom mix. StemDeck runs locally on the user's machine and does not store, cache, or redistribute any downloaded content. It is a free alternative to cloud stem-splitters like Moises and LALAL.AI, with no account, quota, uploads, or subscription required.

StemDeck supports 6-stem separation using Demucs htdemucs_6s with automatic device detection (CUDA on NVIDIA, MPS on Apple Silicon, or CPU fallback). It also offers YouTube and local file import, a DAW-style waveform editor with waveform editing features, stem subset extraction, and a per-stem mixer with volume fader, mute, solo, and monitor options. The application provides state synchronization between the preview mixer and the stems sidebar, live VU meters per stem, and post-gain RMS analysis using Web Audio analyzers.

Additional features include BPM detection, key analysis, loudness measurement, and sample peak detection in dBFS. The macOS and Windows desktop shells are built using Tauri v2, and the frontend is built with vanilla JS and the Web Audio API. StemDeck relies on Python 3.12, FFmpeg, and the Demucs model for stem separation and on-demand lead/backing vocal split via UVR-MDX-NET Karaoke 2 model.