MakeItDrumless

This project allows you to generate drumless tracks from YouTube videos using the SCNet model.

Project Structure

MakeItDrumless/
├── .git/
├── .gitignore
├── README.md
├── requirements.txt
├── checkpoints/ (SCNet model checkpoints)
├── ffmpeg/ (Portable FFmpeg binaries)
├── SCNet/ (SCNet repository - external, for inference)
├── src/
│   ├── __init__.py
│   ├── audio/
│   │   ├── __init__.py
│   │   ├── downloader.py (Handles YouTube audio download)
│   │   └── processing.py (Handles audio mixing and metadata)
│   ├── ffmpeg/
│   │   ├── __init__.py
│   │   └── manager.py (Manages FFmpeg installation)
│   ├── scnet_integration/
│   │   ├── __init__.py
│   │   ├── inference.py (Integrates SCNet inference logic)
│   │   └── repo_manager.py (Manages SCNet repository and checkpoints)
│   └── utils/
│       ├── __init__.py
│       └── spinner.py (Utility for console spinner)
└── temp/ (Temporary files)
└── output/ (Generated drumless tracks)

Setup

1. Clone the repository:

   git clone https://github.com/your_username/MakeItDrumless.git
   cd MakeItDrumless

2. Create and activate a virtual environment:

   python -m venv .venv
   # On Windows
   .venv\Scripts\activate
   # On macOS/Linux
   source .venv/bin/activate

3. Install dependencies:

   pip install -r requirements.txt

4. Initial Setup (FFmpeg and SCNet Model):

   The first time you run main.py, it will automatically:

   • Download and set up a portable FFmpeg (Windows only, otherwise requires manual installation).
   • Clone the SCNet repository into the SCNet/ directory.
   • Check for the SCNet model checkpoint (SCNet-large.th) and configuration file (config.yaml) in the checkpoints/ directory. You will be prompted to manually download the checkpoint if it's missing.

Usage

To generate a drumless track from a YouTube video, run the main.py script with the YouTube URL as an argument:

.venv\Scripts\python.exe src/main.py "<YouTube_URL>"

Example:

.venv\Scripts\python.exe src/main.py "https://www.youtube.com/watch?v=dQw4w9WgXcQ"

Output

The generated drumless track (MP3 format) will be saved in the output/ directory. Separated stems (vocals, bass, other) will be saved in separated_scnet/<video_title>/.

Troubleshooting

• FFmpeg not found: If you are not on Windows, ensure FFmpeg is installed and added to your system's PATH.
• SCNet Checkpoint Missing: Follow the instructions in the console to manually download the SCNet-large.th file.
• Dependency Issues: Ensure all packages in requirements.txt are installed.