K-POP Composition with Generative AI

This research aimed to create a K-POP music composition model with Long Short-Term Memory and analyze K-POP datafication by integrating technology and other fields.K-POP was converted with a graph by Librosa, and it differed from people listening to music with their ears and looking at the data.It shows more relationships between pitch, time, and frequency.The LSTM model trains three data sets: all K-POP genre songs, BLACKPINK's songs, and BTS's songs.The output has a repeated melody when a model trains songs on all K-POP genre songs.It was a repetition of high notes.Conversely, when it focuses on learning one specific group, the output of the model has a wide range of notes.The model trained on a specific group's song generates a song that is closer to K-POP.