A User-Guided Generation Framework for Personalized Music Synthesis Using Interactive Evolutionary Computation

作者：Yanan Wang, Yan Pei, Zerui Ma, Jianqiang Li · 发表于：Proceedings of the Genetic and Evolutionary Computation Conference Companion · 年份：2024 · DOI：10.1145/3638530.3664109 · 被引用次数：6 · 研究领域：Music Technology and Sound Studies、Music and Audio Processing、Evolutionary Algorithms and Applications

The development of generative artificial intelligence (AI) has demonstrated notable advancements in the domain of music synthesis. However, a perceived lack of creativity in the generated content has drawn significant attention from the public. To address this, this paper introduces a novel approach to personalized music synthesis, incorporating a human-in-the-loop generation. This method leverages the dual strengths of interactive evolutionary computation, known for its capturing user preferences, and generative adversarial network, renowned for its capacity to autonomously produce high-quality music. The primary objective of this integration is to augment the credibility and diversity of generative AI in music synthesis, fostering computational artistic creativity in humans. Furthermore, a user-friendly interactive music player has been designed to facilitate users in the music synthesis process. The proposed method exemplifies a paradigm wherein users manipulate latent space through human-machine interaction, underscoring the pivotal role of humans in the synthesis of diverse and creative music.