MusicTheoryKit nears v1

MusicTheoryKit is edging toward something more ambitious than a pile of chord utilities. The project, described this week by its creator as roughly 85 to 90 percent complete in version 0.6, now has working core modules for voice leading, harmony, melody, and an evaluator. The unfinished parts are telling. Rhythm still needs work. MIDI still needs work. That split says a lot about what this kit is trying to become: not a toy that spits out notes, but a system that can represent musical structure cleanly enough to reason about it. (github.com) That matters because MusicTheoryKit did not begin as a generative music headline machine. Its public description on GitHub is much plainer than that. It is a Swift package for modeling practical music theory in code, built as an educational tool that moves from pitches and intervals up through keys, scales, chords, progressions, and beat-based song structure using small, explicit value types. In other words, the project’s center of gravity is representation first. Generation comes later, if it comes at all. (github.com) The companion app around it makes that even clearer. MusicTheoryApp, which sits on top of the kit, is a macOS teaching app with interactive lessons, piano-keyboard exploration, audio playback, and MIDI input support. So the codebase already lives in a world where theory is meant to be handled, heard, and tested, not just stored as abstract symbols. That is why the missing MIDI and rhythm pieces stand out. Once those are stronger, the kit stops being just a way to describe music and starts becoming a more practical bridge between theory rules and actual composition workflows. (github.com) The most useful way to understand the current discussion around the project is not as an argument about whether software can be creative. It is an argument about what music theory is for. In the conversation linked from the card, theory is framed less as a replacement for invention than as a shared language musicians use to describe what they are hearing and trying to do. That is a much narrower claim than the usual AI-music pitch, and a more believable one. A tool that helps you inspect a melody, compare harmonic options, or evaluate a line for voice-leading problems does not need to pretend it wrote the song. It only needs to help the musician talk to the song more clearly. That is also why the specific modules already in place are the important detail. Voice leading is about how individual lines move from one chord to the next with as little awkward motion as possible. Harmony handles the vertical stack. Melody handles the horizontal line. An evaluator implies judgment, or at least scoring, which is the part that turns a theory library into a composition helper. A program can suggest that one progression keeps common tones better, or that one melodic move creates more tension, without claiming that the better choice is universal. That is a very different product from a black box that generates “music” and calls the job done. (online.berklee.edu) So the project’s reported progress is not just a version-number update. It marks the point where the hard conceptual pieces appear to be in place. The remaining work is on timing and interchange with the outside world. That sounds mundane, but it is the concrete bottleneck between a theory engine and something musicians can actually use in a live workflow. MusicTheoryKit can already model keys, scales, chords, progressions, and song structure in code. The next step is making those ideas land on the beat, and then making them leave the app as MIDI. (github.com)