AI Stem Splitter by Oliver Tkach - Version 2.0

A ReaScript for REAPER that uses Demucs to split audio into high-fidelity stems directly in your project.

Features

• 6 stems: vocals, drums, bass, guitar, piano, other
• Quality profiles: fast, balanced, high
• Two-stem mode (target + no_target)
• BPM/key/Hz analysis and project tempo sync
• Automatic track naming, foldering, and colors
• Local processing (no cloud upload)

System Requirements

1. Python 3.10+
   Windows users should enable "Add Python to PATH" during install.
2. FFmpeg
   • Windows: winget install Gyan.FFmpeg
   • macOS: brew install ffmpeg
3. Python dependencies
   pip install demucs soundfile==0.12.1 librosa numpy scipy torchcodec

Installation and Usage (Manual)

1. Download your preferred script:
   • AI Stem Splitter by Oliver Tkach - GUI.lua (recommended)
   • AI Stem Splitter by Oliver Tkach.lua
2. In REAPER, open Action List -> New Action -> Load ReaScript.
3. Select an audio item and run the script.
4. First run may download model files.

ReaPack Installation (Recommended)

Use ReaPack to install and keep scripts updated from this repo.

1. In REAPER, open Extensions > ReaPack > Import repositories...
2. Add: https://raw.githubusercontent.com/tkacholiver/aistemsplitter/main/index.xml
3. Open Extensions > ReaPack > Browse packages
4. Search for AI Stem Splitter by Oliver Tkach - GUI.lua
5. Install and click Apply

Known-good fallback index:

https://raw.githubusercontent.com/tkacholiver/aistemsplitter/reapack-working-2026-02-24/index.xml

Power User Options

• Device: auto, cpu, cuda, mps
• Segment size tuning for low RAM/VRAM systems
• Worker job count tuning

Troubleshooting

• Python not found: reinstall Python and enable PATH integration
• Out-of-memory: lower segment seconds
• FFmpeg missing: install and restart REAPER

Support

• Instagram: https://www.instagram.com/olivertkachmusic
• YouTube: https://www.youtube.com/@olivertkachreactions
• Donate: https://tecito.app/olivertkachmusic