StemDeck releases free, local AI stem separator for musicians and producers

StemDeck is a free, open-source desktop application that separates audio files into up to six stems—vocals, drums, bass, guitar, piano, and other—using local processing on the user's machine.

Users can import local audio files in formats such as MP3, WAV, FLAC, OGG/Opus, MP4, or M4A, or paste a YouTube URL for processing. StemDeck does not store, cache, or redistribute any downloaded content; all processing occurs locally.

The app includes a DAW-style waveform editor with zoom, loop selection, and a seven-lane mixer for adjusting volume, muting, soloing, and monitoring individual stems. It also generates a downloadable mix of selected stems and provides live VU meters per stem.

Additional features include song analysis with BPM detection, key and scale estimation, LUFS measurement, and sample peak detection. Users can cancel ongoing jobs, and the app organizes tracks in a library panel with folder-based organization and search.

StemDeck uses the Demucs model (htdemucs_6s) for six-stem separation and automatically selects the best Torch device (CUDA on NVIDIA, MPS on Apple Silicon, or CPU fallback) for local inference.

The project is released under the Apache-2.0 license and does not accept funding or sponsorships. It positions itself as a privacy-focused, no-cost alternative to commercial cloud-based stem separators such as Moises and LALAL.AI.