DAEJEON, May 8 (Korea Bizwire) — South Korea’s Korea Advanced Institute of Science and Technology (KAIST) has unveiled a cutting-edge artificial intelligence tool designed to assist musicians in the creative process by turning abstract inspirations into musical compositions.
The system, called “Amuse,” was developed by Professor Lee Sung-joo’s research team in KAIST’s Department of Electrical Engineering. It enables users to input a range of stimuli—including text, images, or audio—and translates them into harmonic structures using a large language model (LLM), facilitating real-time musical ideation and composition.
For instance, when given a prompt like “memories of a warm summer beach,” Amuse generates fitting chord progressions aligned with the sentiment, helping composers navigate the early stages of songwriting. The system is also trained to automatically filter out awkward or musically incoherent results, ensuring a higher-quality creative output.

Members of the research team that developed the AI composition technology: From left, Professor Chris Donahue of Carnegie Mellon University, Ph.D. candidate Kim Ye-won and Professor Lee Sung-joo of the Department of Electrical Engineering at KAIST. (Photo provided by KAIST)
In trials with professional musicians, Amuse was praised not merely as a music-generation AI, but as a promising co-creative partner—a tool that complements, rather than replaces, human artistry.
“This project explores how AI can support creativity without infringing on artistic ownership,” Professor Lee said. “Unlike controversial generative models that risk copyright violations, Amuse is designed to preserve the creator’s autonomy while enhancing the collaborative potential between humans and AI.”
The research was recognized internationally, earning a Best Paper Award—granted to the top 1% of submissions—at ACM CHI 2025, a prestigious human-computer interaction conference held this year in Yokohama, Japan.
Kevin Lee (kevinlee@koreabizwire.com)