Amuse: A New AI Tool for Interactive Songwriting That Supports Musicians' Creative Flow

The world of music composition is poised to welcome a new collaborator: Amuse, an AI-based songwriting assistant developed by researchers at KAIST and Carnegie Mellon University (CMU). This innovative system aims to transform the way music creators brainstorm, overcome writer's block, and experiment with different musical directions. Unlike traditional generative AI models, Amuse focuses on enhancing the human creative process rather than taking over it entirely. Amuse operates by converting various forms of inspiration—such as text, images, and audio—into harmonic structures known as chord progressions. For instance, if a user inputs a phrase like "memories of a warm summer beach," Amuse will automatically suggest chord sequences that evoke the desired mood and atmosphere. This capability helps composers translate abstract concepts into concrete musical elements, streamlining the songwriting process. One of the key differentiators of Amuse is its interactive and flexible approach. The system is designed to respect the composer's creative flow by allowing users to modify and integrate AI-generated suggestions seamlessly. This method stands in contrast to traditional AI models that often generate music independently, leading to results that may not align with the creator's vision. By facilitating a more collaborative and adaptive interaction, Amuse aims to enhance creativity and maintain the artist's control over their work. The core technology behind Amuse involves a two-pronged generative approach. First, a large language model transforms the user's input into music code. This initial step creates a foundation for the chord progressions. Then, another AI model, trained on real music data, filters out any results that might seem awkward or unnatural using rejection sampling. This dual-process ensures that the generated chord progressions are both inspired by the user's input and grounded in practical, harmonically sound compositions. To validate the system's effectiveness, the research team conducted a user study involving professional musicians. The participants were tasked with using Amuse to create music and then evaluated the tool's performance. The results were highly positive, indicating that Amuse has significant potential as a co-creative AI companion. Musicians found the AI suggestions helpful for sparking new ideas and exploring novel musical directions while maintaining their own artistic influence. Professor Sung-Ju Lee of KAIST's School of Electrical and Electronic Engineering emphasized the importance of ethical considerations in the development of generative AI. He noted that recent advances in AI have raised concerns about copyright violations and the lack of alignment with creator intentions. In response, the Amuse team focused on understanding the true needs of music creators and designed a system that prioritizes the creator’s initiative. Lee stated, "Amuse is an attempt to explore the possibility of collaboration with AI while maintaining the initiative of the creator, and is expected to be a starting point for suggesting a more creator-friendly direction in the development of music creation tools and generative AI systems in the future." The research, co-led by Ph.D. student Yewon Kim from KAIST and CMU Professor Chris Donahue, was presented at the Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems. Their paper highlights the potential of Amuse in both academic and industrial settings, underscoring the importance of user-centric design in AI-powered creative tools. Industry insiders are enthusiastic about the possibilities that Amuse offers. They see it as a significant step forward in the field of human-AI collaboration, particularly in music. By focusing on the creative flow and flexibility, Amuse addresses some of the key challenges associated with AI in content creation, such as the preservation of artistic individuality and the prevention of copyright issues. Companies that specialize in music technology, like Spotify and Apple Music, could benefit greatly from integrating similar co-creative AI tools into their platforms, potentially revolutionizing the way music is created and shared. KAIST, one of Asia’s premier institutions for science and technology, is committed to advancing research at the intersection of AI and human creativity. Their collaboration with CMU, a renowned leader in AI and machine learning, exemplifies the synergy of top-tier academic institutions working towards innovative solutions in the tech industry. Amuse is not just a tool but a symbol of the future where AI and human creativity can coexist harmoniously, enhancing each other's capabilities without overshadowing the unique human touch.