An automated mashup system with drum-bass stem swapping

Music mashups have emerged as a compelling form of art, merging elements from various songs to create new auditory experiences. Our research automated the mashup generation process and analyzed listener preferences to understand what contributes to a mashup's popularity. Leveraging modern source separation technologies, we enabled the automatic creation of mashups from any two songs. Using three mixing approaches—drum swapping, bass swapping, and combined drum and bass swapping—we generated 270 mashups for evaluation. These approaches were tested with advanced beat tracking and harmonic alignment techniques to ensure musical coherence. Systematic listening tests were conducted with 50 survey participants to gauge preferences. The analysis revealed low correlations in preferences across the different approaches, indicating that each created a unique listening experience. Coupled with a comprehensive analysis of rhythm and harmony, our research sheds light on the qualities that make a mashup appealing and provides guidelines for developing more effective automated tools to assist music enthusiasts in creating popular and musically satisfying mashups.