AI Music Generation

Enhancing AI Music Generation with Multi-Module Neural Networks A Novel Approach to Chord, Rhythm, and Pitch Synthesis

Table of Contents

• Overview
• Features
• Installation
• Usage
• License

📌 Overview

• Automatic music generation plays a crucial role in generating creative compositions autonomously, facilitating applications in various fields, including entertainment and education.
• The current approaches often struggle to:
  • Capture temporal dependencies across sequences.
  • Maintain harmonic coherence.
  • Generate dynamic rhythmic patterns effectively.
• This project introduces a Multi-Module Neural Network (MNN) for automatic music composition, combining several specialized modules.
• The modules are:
  • 🎼 Temporal Graph Dilated Convolution (TGDC): Captures chord progression dependencies.
  • 🥁 Rhythm Generator (MobileNet-TCN): Creates expressive rhythmic patterns.
  • 🎹 Pitch Generator (Sparse Transformer): Generates coherent pitch sequences aligned with rhythm.
  • 🎶 Fusion & Post-Processing: Merges rhythm and pitch while refining timing, velocity, and harmonic structure.
• The result is a highly expressive, scalable music generation system that outputs MIDI files or audio-ready compositions.

🚀 Features

• Multi-module neural architecture for chords, rhythm, and pitch.
• High accuracy and harmonic coherence.
• 98.98% Accuracy and 98.13% F1 Score in experimental results.
• Support for MIDI-based dataset training.
• Automatic evaluation (Accuracy, Precision, Recall, F1, BRA, CTR).

🔧 Installation

Clone repository

https://github.com/Athira1979/multi-module-music-gen.git
cd multi-module-music-gen

Create virtual environment

python -m venv .venv
python  .venv\Scripts\activate # activate in windows

Install requirements

pip install -r requirements.txt

Usage

Train the model

python main.py --mode train

Evaluate the model

python main.py --mode eval

License

This project is licensed under the MIT License - see the LICENSE file for details.