StemDeck ghcr.io/stemdeckapp/stemdeck:edge https://github.com/stemdeckapp/stemdeck/pkgs/container/stemdeck bridge sh false https://github.com/stemdeckapp/stemdeck/issues https://github.com/stemdeckapp/stemdeck https://raw.githubusercontent.com/stemdeckapp/stemdeck/main/templates/stemdeck.xml https://raw.githubusercontent.com/stemdeckapp/stemdeck/main/README.md https://raw.githubusercontent.com/stemdeckapp/stemdeck/main/desktop/src-tauri/icons/icon.png http://[IP]:[PORT:8000]/ MediaApp:Music MediaApp:Video Tools: stems demucs vocal remover karaoke isolation multitrack youtube audio separation Optional: install the Unraid "Nvidia Driver" plugin and set Extra Parameters to --runtime=nvidia for GPU acceleration. Runs on CPU without a GPU. StemDeck splits any song into isolated stems (vocals, drums, bass, guitar, piano, other) and plays them back in a DAW-style multitrack player. Paste a YouTube URL or upload a file, get separated tracks you can mute, solo, loop, slow down, and pitch-shift for practice and transcription. Powered by Demucs. Data: processed tracks and downloads live under /app/jobs; Demucs model weights and the torch cache live under /cache. Map both to persistent appdata paths so nothing re-downloads on update. The library is persistent and user-managed on Unraid (STEMDECK_PERSIST_LIBRARY=1 is set by default); tracks are never auto-deleted. GPU (optional): install the Unraid "Nvidia Driver" plugin, then set Extra Parameters to --runtime=nvidia and keep NVIDIA_VISIBLE_DEVICES / NVIDIA_DRIVER_CAPABILITIES. StemDeck auto-detects CUDA and uses it, processing several times faster. Without a GPU it runs on CPU with no extra configuration. Note: very new Blackwell cards (RTX 50 series) may fall back to CPU until a newer CUDA build ships. Split songs into stems and play them in a DAW-style multitrack player. Self-hosted, GPU-accelerated with the Unraid Nvidia Driver plugin. 8000 /mnt/user/appdata/stemdeck/jobs /mnt/user/appdata/stemdeck/cache 1 1200 3 99 100 all all