A Study on the Production of Real-Time Algorithmic Composition System Using the VQ-VAE Algorithm of OpenAI Jukebox

Abstract

Algorithm composition user production environments using a computer can be used only by waiting for data calculated using predefined functions based on variables declared through a certain level of coding or programming, making it difficult to produce works through algorithmic composition in real time. Algorithm composition is possible using audio data recorded in real time through OpenAI s Jukebox, and when organic combination with various programs used in interactive media art production environments such as Max/MSP is achieved, it is accidental with minimal intervention. Jukebox is an AI that takes musical quality, consistency, audio sample length, and ability to control artists, genres and lyrics to the next level, but it is quite different from human-made music. However, with the least human intervention that could not be done in other AI programs, real-time algorithm works can be produced by chance, which will allow them to more be free in a tool called a program for creating real-time multimedia works and produce a wide range of results, not limited ones.