Deep Learning-Based Variation in Music Composition

This thesis explores how Artificial Intelligence (AI) can be integrated into musical practice by investigating how a system that generates variations on a composer's musical phrases can support their exploratory process during composition. This was achieved through the design and development of Rhapsody Refiner, a music variation system, created in collaboration with sixteen practising musicians. Through this work, the thesis demonstrates the utility of AI music variation as a compositional tool and highlights how it supports creative ownership. Furthermore, this work argues that musicians should be more actively included in the design of AI-based music technology.