MelodicNet is a generative AI application tailored to aid music producers in melody creation. By inputting an "inspiration" MIDI file, the app generates new MIDI files that echo elements of the original composition. While the current output may require further editing before integration into a final composition, MelodicNet serves as a tool for sparking creativity and exploration. Rather than providing fully structured melodies, it excels in generating unique melodic phrases, offering users a starting point for musical experimentation and development.
This project was developed as part of the capstone design course for my undergraduate degree in computer engineering.
The deployed model is trained on a dataset of 500 hip-hop MIDI files. View the backend directory to train your own model.