KAIST Amuse AI: Music Composition Tool Released

[헤럴드경제=구본혁 기자] Domestic researchers have developed artificial intelligence (AI) technologies such as fellow writers who help music creation.

KAIST announced on the 7th that Lee Sung -joo, a professor of electrical and electronic engineering, developed AI -based music creative support system Amuse. The findings were awarded the Best Thesis Award, which was awarded only to the top 1%of all papers in CHI, the world’s most prestigious international conference in human-computer interaction.

The Amuse System enters a variety of inspiration, such as text, images, and audio, and converts it to Mars structure (code progress) to support composition.

For example, when a user enters a phrase, an image, or a sound clip, such as ‘Memory of Warm Summer Beach’, Amuse automatically generates and suggests code progression that suits the inspiration.

Unlike the existing AI, Amuse is differentiated in that it naturally encourages creative search through the interaction method that respects the user’s creative flow and flexibly integrates and modifies the proposal of AI.

The core technology of the Amuse System is a hybrid creation method that creates a music code that is input to the prompt in inspiration of a large language model to create a music code that suits it, and the AI ​​model who has learned actual music data naturally reproduces the two methods that are combined through the unnatural or awkward (rejection sampling) process.

The team conducted user studies for actual musicians, and was evaluated that Amuse is not just a music creation AI, but a creative partner (Co-CREATAI) in collaboration with people.

Professor Lee Sung -joo said, “Amus is an attempt to explore the possibility of collaboration with artificial intelligence while maintaining the initiative of the creator, and is expected to be a starting point for the development of music creative tools and generated AI systems in the future.”

[email protected]