Introducing musicpy: A Python Library for Music Generation and Theory‑Driven AI Composition

Ever wondered how to write music with code? musicpy, a Chinese‑authored Python library, allows you to express notes, rhythms, and other musical elements using a compact, theory‑driven syntax that can be exported as MIDI files. The library assumes some basic music‑theory knowledge, but experienced users can quickly get started by following the author’s Wiki tutorial.

Installation

Open a command prompt in the musicpy folder and run pip install -r requirements.txt to install dependencies (if you have already installed the library via pip install musicpy , this step is optional). Then import the library with from musicpy import * .

Watch the author’s introductory video on Bilibili: https://www.bilibili.com/video/BV1754y197a9/ . Detailed data structures, basic syntax, and usage tutorials are available in the Wiki: https://github.com/Rainbow-Dreamer/musicpy/wiki .

The core classes in musicpy are note (note), chord (chord), and scale (scale). These classes are designed as pure numeric values, enabling them to participate in mathematical operations such as linear algebra or discrete mathematics. Consequently, you can build comprehensive music‑theory algorithms and conduct analytical research on music.

Because the data structures are purely numeric, experimental music styles—serialism, aleatoric, minimalism, and others—can be rigorously composed using musicpy. The library also supports classical, jazz, and pop music composition.

The changelog ( Changelog.txt ) records frequent updates, typically every two to three days.

Motivation

The author created musicpy for two main reasons: (1) to represent music composition through theory‑based code rather than raw MIDI data, dramatically reducing the amount of code needed; and (2) to provide an interface for AI to compose music by understanding music theory instead of relying on large‑scale deep‑learning models.

By teaching AI explicit theory rules, the library enables AI to generate music with human‑like reasoning, adjustable style preferences, and genuine musical understanding, avoiding the mechanical feel of data‑driven models.

Conclusion

musicpy is available on PyPI; install it with pip install musicpy . Development began in October of the previous year and continues with regular tutorial videos and documentation. The GitHub repository is https://github.com/Rainbow-Dreamer/musicpy .