This thesis studies a methodology that applies the diffusion model to melody creation. The research focuses on developing techniques to precisely control pitch and rhythm density during the melody generation. These elements are pivotal in defining a m...
This thesis studies a methodology that applies the diffusion model to melody creation. The research focuses on developing techniques to precisely control pitch and rhythm density during the melody generation. These elements are pivotal in defining a musical piece’s emotional and stylistic identity. The objective is to enhance the emotional depth and structural complexity of generated melodies through strategic manipulation of pitch and rhythm. The primary aim of this study is to ensure that the generated melodies align cohesively with the existing melodic context. This alignment provides a seamless and natural musical experience for listeners. The research presents a novel approach that leverages diffusion models to improve melodies’ continuity and coherence. Secondly, the thesis explores effective pitch control in melody generation. Recognizing pitch as a critical component in conveying emotions and setting the mood of a piece, this research delves into methods for adjusting pitch based on user inputs. Lastly, the study addresses the control of rhythm density. Rhythm, a fundamental aspect in dictating a composition’s flow and energy, is manipulated to diversify and enhance the complexity of melodies. This approach aims to broaden the expressive capabilities of music generated through this novel methodology.