Audio Stem Splitter

A simple command-line tool that uses Meta's Demucs AI model to separate mixed audio files (like MP3s or WAVs) into individual stems:

• Vocals
• Drums
• Bass
• Other (Melody/Instruments)

Prerequisites

1. Python 3.8+
2. ffmpeg: Required for handling various audio formats.
   • macOS: brew install ffmpeg
   • Ubuntu/Debian: sudo apt install ffmpeg
   • Windows: Download from gyan.dev or use winget install ffmpeg

Installation

1. Clone or download this repository.
2. (Optional but recommended) Create a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows, use `venv\Scripts\activate`

3. Install the required Python libraries:

   pip install -r requirements.txt

Usage

Run the script from your terminal, providing the path to your audio file.

Basic Usage

python split_stems.py path/to/your/song.mp3

By default, the separated stems will be saved into a separated folder in your current directory.

Advanced Usage

You can specify a custom output directory and choose which Demucs model to use:

python split_stems.py path/to/your/song.mp3 -o /path/to/output_folder -m htdemucs_ft

Available Models (-m flag):

• htdemucs (Default)
• htdemucs_ft (Fine-tuned, slightly better quality but takes longer)
• htdemucs_6s (Separates into 6 stems: vocals, drums, bass, other, piano, guitar)
• mdx_extra
• mdx_extra_q (Quantized, faster but slightly lower quality)

Note: The first time you run the script, Demucs will download the pre-trained neural network weights. This may take a few minutes depending on your internet connection.