Skip to main content

StemX Overview

AI-powered real-time stem separation for producers and audio engineers.

StemX uses a BS-RoFormer neural model to separate any audio into six distinct stems: Vocals, Drums, Bass, Guitar, Piano, and Other (synths, strings, etc.).

StemX runs entirely on your machine, with no upload. Live mode separates audio inside your DAW as it plays; File mode loads and processes whole files. Separation is GPU-accelerated, and on Apple Silicon it runs faster than real time.

The Six Stems

VOXVocals & Voice
DRMDrums & Percussion
BASBass & Sub
GTRGuitar
PNOPiano & Keys
OTHOther Instruments

Key Features

  • Real-time in the DAW

    Live mode separates incoming audio as it plays, for monitoring and on-the-fly stem control.

  • Two Operating Modes

    Live mode for DAW input, File mode for loading and exporting audio files.

  • Per-Stem Controls

    Independent volume, mute, solo, and pan for each of the six stems.

  • GPU Acceleration

    CUDA (NVIDIA), DirectML (Windows), and CoreML (Apple Silicon) support.

  • Stem Export

    Export individual stems as WAV, FLAC, or MP3 files for further processing.

  • Fully offline

    All processing happens locally. Your audio never leaves your machine.

Technical Specifications

SAMPLE RATES44.1-96kHz
STEMS6
AI MODELBS-RoFormer
GPU ACCELCUDA / DirectML / CoreML
FORMATSVST3 / AU / Standalone
EXPORTWAV / FLAC / MP3