MIDI Bass Generation
Music generation has been an active research area for a long time. Researchers in the field of music information retrieval (MIR) have proposed different models and algorithms that lay a successful foundation for further developments. Currently, research and industry mainly focus on automatic music composition and generation from scratch. There are already different music generation apps and programs available. However, they require specific inputs such as genre and tempo to create new music in seconds. Although results are suitable for background music in films or if license-free music is required, the generated music lacks creativity and thus, cannot compete with human-made music. Therefore, the goal of this bachelor thesis is to minimize the research gap between generated music and human creativity by developing a program that extends a piano composition with a generated bass line. We aim to generate a bass line, which matches the piano MIDI input, based on deep neural network architectures.