JSBach - Bach Music Generation with AI

AI model that generates Bach-style music compositions using machine learning.

Project Structure

JSBach/
├── scripts/                    # Data collection scripts
│   ├── scrape_bach_midi.py    # Scraper for jsbach.net
│   └── scrape_bachcentral.py  # Scraper for bachcentral.com
│
├── data/
│   ├── raw/                   # Raw downloaded MIDI files
│   │   ├── jsbach_net/       # ~257 files from jsbach.net
│   │   └── bachcentral/      # Files from bachcentral.com (run scraper)
│   │
│   ├── final/                 # Merged & deduplicated dataset (future)
│   └── processed/             # Piano-converted training data (future)
│
└── notebooks/                 # Jupyter notebooks for exploration (future)

Setup

# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate

# Install dependencies
pip install requests beautifulsoup4 pretty_midi numpy pandas music21

Data Collection

Download from jsbach.net (already done - 257 files)

cd scripts
python scrape_bach_midi.py

Download from BachCentral

cd scripts
python scrape_bachcentral.py

Current Status

✅ Phase 1: Data Collection (In Progress)

• jsbach.net: 257 MIDI files downloaded
• bachcentral.com: Ready to download

⏳ Next Steps:

• Complete bachcentral download
• Deduplicate and merge datasets
• Convert all instruments to piano
• Create training sequences
• Build and train model

Dataset Summary

Current Collection (jsbach.net - 257 files):

• Goldberg Variations: 33 files
• English Suites: 48 files
• Solo Cello: 36 files
• Solo Violin: 31 files
• Organ Works: 32 files
• Well-Tempered Clavier (partial): 49 files
• Art of Fugue: 4 files
• And more...

Target: 500-600 unique Bach MIDI files for training