Might 7, 2025
The GIST Editors' notes
This text has been reviewed in keeping with Science X's editorial course of and insurance policies. Editors have highlighted the next attributes whereas guaranteeing the content material's credibility:
fact-checked
trusted supply
proofread
Amuse, a songwriting AI collaborator for music composers
Wouldn't it’s nice if music creators had somebody to brainstorm with, assist them once they're caught, and discover totally different musical instructions collectively? Researchers at KAIST and Carnegie Mellon College (CMU) have developed AI know-how much like a fellow songwriter who helps create music.
The work is revealed in Proceedings of the 2025 CHI Convention on Human Elements in Computing Methods.
The system developed by Professor Sung-Ju Lee's analysis crew, Amuse, is an AI-based system that converts numerous types of inspiration reminiscent of textual content, photos, and audio into harmonic buildings (chord progressions) to assist composition.
For instance, if a consumer inputs a phrase, picture, or sound clip reminiscent of "recollections of a heat summer season seaside," Amuse mechanically generates and suggests chord progressions that match the inspiration.
Not like current generative AI, Amuse is differentiated in that it respects the consumer's artistic stream and naturally induces artistic exploration by means of an interactive technique that enables versatile integration and modification of AI ideas.
The core know-how of the Amuse system is a technology technique that blends two approaches: a big language mannequin creates music code primarily based on the consumer's immediate and inspiration, whereas one other AI mannequin, educated on actual music knowledge, filters out awkward or unnatural outcomes utilizing rejection sampling.
The analysis crew performed a consumer examine concentrating on precise musicians and evaluated that Amuse has excessive potential as a artistic companion, or a Co-Inventive AI, an idea during which folks and AI collaborate, relatively than having a generative AI merely put collectively a track.
The paper, during which a Ph.D. scholar Yewon Kim and Professor Sung-Ju Lee of KAIST College of Electrical and Digital Engineering and Carnegie Mellon College Professor Chris Donahue participated, demonstrated the potential of artistic AI system design in each academia and trade.
Professor Sung-Ju Lee stated, "Current generative AI know-how has raised issues in that it instantly imitates copyrighted content material, thereby violating the copyright of the creator, or producing outcomes one-way whatever the creator's intention. Accordingly, the analysis crew was conscious of this pattern, paid consideration to what the creator really wants, and targeted on designing an AI system centered on the creator."
He continued, "Amuse is an try and discover the potential for collaboration with AI whereas sustaining the initiative of the creator, and is anticipated to be a place to begin for suggesting a extra creator-friendly course within the improvement of music creation instruments and generative AI techniques sooner or later."
Extra info: Yewon Kim et al, Amuse: Human-AI Collaborative Songwriting with Multimodal Inspirations, Proceedings of the 2025 CHI Convention on Human Elements in Computing Methods (2025). DOI: 10.1145/3706598.3713818
Offered by The Korea Superior Institute of Science and Expertise (KAIST) Quotation: Amuse, a songwriting AI collaborator for music composers (2025, Might 7) retrieved 7 Might 2025 from https://techxplore.com/information/2025-05-songwriting-ai-collaborator-music.html This doc is topic to copyright. Aside from any honest dealing for the aim of personal examine or analysis, no half could also be reproduced with out the written permission. The content material is supplied for info functions solely.
Discover additional
An energy-efficient text-to-audio AI 22 shares
Feedback to editors
