Music often serves as a channel for human expression and creativity, especially as a conduit for emotion. Music theory is an art surrounding music that dictates how chords interplay and what melodies work. Composing original music can be a very difficult task, particularly for those without formal training in music theory or composition. To address this challenge, our proposed project aims to give users the ability to generate personalized musical compositions effortlessly. Utilizing traditional computer programming methods and machine learning algorithms, the program will accept user-provided input, consisting of one or more midi files to base the song on. The program's core functions include training a model using an optimizer training loop on a large-scale musical dataset, preprocessing data into multiple data formats for comprehensive use, generating compositions using a decoder-only transformer, and exporting these compositions into MIDI files. The proposed approach contributes to creativity in programming, and it holds practical uses for users seeking innovative tools for musical inspiration.