HarmonyHub is an open-source project that integrates artificial intelligence with music education. By leveraging models like MIDI-GPT and TransformerXL, it generates customized musical exercises and offers adaptive feedback to enhance the music learning experience.
Developed as part of Google Summer of Code 2025 under the International Neuroinformatics Coordinating Facility (INCF), this project aims to make music education more accessible, interactive, and personalized.
HarmonyHub is built to support learners and educators through:
- Personalized generation of practice material based on current skill level
- Real-time feedback on rhythm, pitch, and timing
- Dynamic adaptation of difficulty and content as users improve
- Visualization of generated music in MIDI and sheet formats
- MIDI-GPT: A transformer-based model used for generating expressive and coherent MIDI sequences.
- TransformerXL: Used for modeling long-term structure in user practice patterns.
- Interactive Interface: Allows users to input, listen to, and evaluate generated music.
- Performance Evaluation: Tracks user progress and adapts future exercises accordingly.
- AI-generated custom exercises and melodies
- Feedback loop based on user input and performance
- Modular architecture for adding new instruments or training models
- MIDI support for both input and output
- Sheet music rendering (planned feature)