tSA[track Select Assistant]

Yuga Kobayashi
Ryo Nishikado
Ryo Hasegawa


Many DJs, when they perform in clubs, feel the atmosphere of the place while DJing. However, if there is an AI that has been trained with the data of its own music, it will only select the next song based on the musical characteristics of the current song. There can be some expression that human DJs can learn from this. In response to this question, we developed the prototype tSA [track Select Assistant], which extracts features from the DJ’s own song library using a genre discrimination model, and suggests five candidates for the next song based on parameters such as mood and song similarity. In this project, we implemented the algorithm of model generation, visualization, and system construction using programming and tools.