AI music generation is evolving from a novelty button into a responsive creative partner. Rather than simply pressing a button to generate a complete song, researchers at UC San Diego are developing AI systems that musicians can play with, shape, and control in real time, fundamentally changing how artists think about AI's role in music creation.

What's Wrong With Today's AI Music Generators?

Current AI music tools follow a simple formula: you write a text prompt, and the system returns a finished song ready to use. While convenient for casual projects, this approach frustrates working musicians who need creative control. Tornike Karchkhadze, a graduating Ph.D. student in music at UC San Diego, explained the disconnect between what these tools offer and what musicians actually need.

"Normally, you write a prompt and get a song back, a fully composed, ready-to-use song. That's fine if you want to make a surprise for your friend on their birthday. But if you are a musician, you need more control. You have your own taste and your own vision of how your music must sound," said Karchkhadze.

Tornike Karchkhadze, Ph.D. Student in Music, UC San Diego

Karchkhadze's research has tackled this limitation head-on. Over several years, he developed AI systems for accompaniment generation, multi-channel music creation, and even interpreting experimental graphic notation. His most recent breakthrough is a real-time human-AI co-performance system that allows a musician to plug in an instrument, play, and receive AI-generated accompaniment that responds to their performance.

How Are Researchers Building AI That Listens and Responds?

The technical challenge is significant. In live musical collaboration, an AI system must respond to what a musician is playing while also anticipating where the performance might go. Human musicians develop this skill naturally through years of practice, but replicating it in machine learning requires sophisticated design. Karchkhadze's real-time system represents a step toward a suite of AI tools that can participate in live musical practice, not just produce music offline.

Karchkhadze's journey also reveals how AI music research is evolving. Trained as a musician rather than a computer scientist, he taught himself to code and build AI systems through practice. That dual perspective shapes his work: he understands both the creative needs of performers and the technical constraints of machine learning. After graduation, he will join Apple's audio research group to work on video-to-audio generation.

Other UC San Diego researchers are tackling different aspects of AI music. Mingyang Yao, who recently graduated with a double major in mathematics-computer science and cognitive science, focused on how AI models learn musical structure and style. Large music-generation models typically rely on massive datasets, but when the goal is to generate music in the style of a specific composer, available data may be limited. Yao explored whether an AI model can first learn broad musical knowledge, then adapt that knowledge to a specific style using limited data.

In one project, Yao pre-trained a symbolic music model on a broad collection of classical, folk, and popular music, then fine-tuned it on the work of composers such as Bach and Mozart. The results showed strong style adaptation using fewer than 300 pieces from a given composer, comparing favorably with much larger models trained on more data. Yao later turned to helping AI systems understand harmony in written music, creating a model that makes decisions step by step, beginning with the features it is most confident about.

Steps to Develop More Playful and Responsive AI Music Systems

• Include Musicians Early: Involve artists throughout the development of AI systems, not just as end users. Zachary Novack, a graduating Ph.D. student in computer science, emphasized that creative AI tools work best when artists are involved from the beginning of the design process.
• Prioritize Real-Time Interaction: Design generative music systems that move beyond one-shot song generation and become more like instruments, tools artists can play with, shape, challenge, and even creatively misuse.
• Optimize for Local Performance: Shrink and speed up models so they can run locally, respond to controls such as pitch and volume, and become part of a performance setup without requiring cloud processing.

Novack's philosophy runs through his research. As co-creator of Presto, a model for accelerating music generation, he explored how to make AI music systems faster and more interactive. In one experimental project, Novack worked on making open-source generative music models responsive enough for real-time interaction. One unusual but revealing result came from a model trained on whale sounds and used in a performance with cello. The system functioned almost like a generative delay effect, responding to the performer in ways that were strange, imperfect, and creatively suggestive.

"The idea of 'press a button and we'll generate the song' is boring to me. It's a toy, but it's not actually fun. Nobody is going to beat the gigantic companies with their gigantic song generators, but who wants to?" said Novack.

Zachary Novack, Ph.D. Student in Computer Science, UC San Diego

What Role Does Attribution Play in AI Music's Future?

While UC San Diego researchers focus on making AI music tools more creative and responsive, the broader music industry is grappling with how to ensure artists benefit from AI. Warner Music Group recently acquired Sureel AI, an attribution startup that traces how AI models use artists' work in training and generation.

Sureel's technology creates what the company calls "AI DNA" for every work, breaking it into component parts and tracing how AI models use those elements. The platform also delivers intellectual property provenance, audit and compliance reporting, model optimization, and a growing name, image, and likeness (NIL) attribution suite that tracks how artist voices, likenesses, and performance identities are used in AI training and generation, including voice clones, AI-generated avatars, and style replication.

"Rightsholders deserve to know how AI interacts with their work, and to share fairly in the value it creates," said Dr. Tamay Aykut, Chief Executive Officer and Founder of Sureel AI.

Dr. Tamay Aykut, Chief Executive Officer and Founder, Sureel AI

The Sureel acquisition reflects Warner Music Group's broader strategy in AI music. In November 2025, the company struck a licensing deal with AI music generator Suno and settled its copyright lawsuit against the company. WMG also settled its lawsuit with Udio and struck a licensing deal with the company for a next-generation AI music platform coming in 2026.

These developments signal a maturing AI music ecosystem. Rather than viewing AI as a threat to be blocked, the music industry is increasingly working to integrate AI tools while ensuring artists and rightsholders maintain control and receive fair compensation. The UC San Diego research demonstrates that the most interesting future for AI music lies not in replacing musicians, but in creating tools that enhance their creativity and give them greater control over the final product.