A big part of music is chords. These are multiple notes played together at the same time, and if you put the right ones together at the right time, they sound nice. But sometimes this is hard.

What we are working on is an ML optimal chord generator.

  • It will take in chord data from a person playing on a MIDI keyboard, and then from that history, if a person needs help deciding where the music/chords should go, they can use our tool.
  • It will generate different “optimal” chords based on how you want the music to sound. So it will use the previous chords that you play to help you pick the next chord if you get stuck, meaning that your music will always sound nice! And “nice” is sometimes different for different people, so there will be multiple options!

We plan to use RAG and AI agents to essentially create an AI agent who is a “music theory expert” and use LangChain or LangGraph to create automated systems to find optimal chords.

Minimum Viable Product: A simple command-line tool that, when it runs, takes in MIDI input and prints out a few optimal chords depending on the desired feel.

Project Leads