Human–AI Co-Creation in Melody Writing: A MIDI-Native Pilot on Flow, Authorship, and Control

Human–AI Co-Creation in Melody Writing

Pilot for my Human-AI Co-Creation Project

This project explores human–AI melodic collaboration using symbolic sequence models such as the Music Transformer and RNN-based anticipation networks.
It includes experiments for real-time melody co-writing and analysis of creative agency between human and machine.

Features

• Symbolic melody generation and continuation using Transformer and RNN models
• Real-time co-creation pipeline (run_anticipation.py)
• Model comparison and evaluation tools
• Ready for ISMIR reproducibility (Python 3.8+)

Model

• stanford-crfm/music-small-800k
• Autoregressive transformer trained on 800k MIDI sequences
• Hosted via HuggingFace Transformers

Data and artifacts

• Project materials and study artifacts: https://doi.org/10.17605/OSF.IO/P3HQG

Quick Start

1. Clone the repository

   git clone https://github.com/your-username/your-repo-name.git
   cd your-repo-name
   

2. Create and activate a Python 3.8 environment python3.8 -m venv venv source venv/bin/activate # macOS/Linux venv\Scripts\activate # Windows

3. Install dependencies:

   pip install -r requirements.txt
   

4. Run generation:

   python run_anticipation.py --input my_melody.mid --output output.mid

Folder overview anticipation/ # Core model files utils/ # Helper functions generate.py # Model inference script run_anticipation.py # Main experiment runner requirements.txt # Dependencies README.md # Project documentation LICENSE # License information