StemDeck: Free Open-Source 6-Stem AI Audio Separator
Local Six-Stem Separation with Demucs
StemDeck is a new open-source desktop application designed for local audio stem extraction. The software splits complete audio recordings into six discrete tracks. Specifically, it isolates vocals, drums, bass, guitar, piano, and other accompaniment elements. Under the hood, the system uses Meta AI’s Demucs architecture, particularly the htdemucs_6s model. Consequently, producers and musicians can separate tracks without uploading files to remote servers.
Furthermore, the utility functions entirely offline once installed. Cloud-based isolation services often require recurring monthly subscriptions. Similarly, remote platforms can raise serious privacy concerns with unreleased audio. Therefore, StemDeck eliminates cloud dependence by processing everything on your computer. In addition, the software contains zero tracking scripts or telemetry.
StemDeck brings completely offline six-stem extraction and multitrack playback directly to desktop workstations.
Modern Architecture and Hardware Acceleration
The desktop software packages a lightweight Tauri shell with a private Python backend. Because local neural processing demands computer resources, the engine supports hardware acceleration. For example, macOS systems run Demucs using Apple Silicon Metal Performance Shaders (MPS). Meanwhile, Windows setups can utilize NVIDIA CUDA acceleration for fast computation. Standard CPU fallback is also available if dedicated hardware is absent.
In addition to standard six-stem splitting, StemDeck offers extra processing choices. For instance, the app supports an optional lead and backing vocal separation mode via UVR-MDX-NET. As a result, remixers can quickly isolate vocal harmonies from the main hook.
Integrated DAW Mixer and Audio Analysis
StemDeck goes beyond simple command-line tools by featuring an interactive multitrack mixer. After separation finishes, waveforms render directly in a browser-based environment. Furthermore, each stem channel features individual solo, mute, volume faders, and live audio metering. Users can also loop short song sections to practice complex riffs.
Additionally, the program provides automated music analysis. For instance, it automatically detects song tempo, musical key, scale, and integrated LUFS levels. Producers can also adjust audio playback speed and transpose pitch without leaving the program. Subsequently, users can export individual WAV files or bounce a customized backing track mix.
Features
- Open-source local stem extraction powered by Meta Demucs htdemucs_6s model
- Six-stem separation: vocals, drums, bass, guitar, piano, and other instruments
- Optional lead and backing vocal separation using UVR-MDX-NET
- Hardware acceleration on Apple Silicon MPS and NVIDIA CUDA GPUs
- Interactive multitrack mixer with waveform display, solo, mute, and level controls
- Built-in music analysis including BPM, key, scale, and integrated LUFS metering
- Pitch transposition and variable tempo playback controls for practice
- Zero telemetry, fully offline processing, and open-source Apache 2.0 license
Price
Intro: Free
Regular: Free
More info here: StemDeck | StemDeck
Pros
- Completely free, open-source, and offline with zero telemetry
- Extracts six distinct stems including guitar and piano via Demucs
- Integrated multitrack DAW-style mixer with key and BPM detection
Cons
- Requires capable GPU hardware for faster separation speeds
- Initial launch requires downloading model weights and runtimes
More info: StemDeck | StemDeck
HIT OR SHIT Indicator
💩
Indication by Noizefield
Average Indication by Readers
🚀
What do you think?
Hit or Shit? Please rate from 1 (💩) to 10 (🚀)
✅ You already voted. Thanks!
Check these out!
