Reimagining Korean Traditional Music with AI: A Journey into Gugak’s Digital Future

Introduction

“How can AI honor centuries-old improvisational traditions?”Gugak (Korean traditional music) is rooted in deep emotional expression, flexible rhythms, and nuanced ornamentation. Unlike Western classical music, Gugak thrives on spontaneity and human sensitivity—qualities that challenge current AI systems. This project explores the fusion of cultural heritage and emerging technology by developing AI-assisted tools for Gugak production, transformation, and global accessibility.

This work is personal to me. As someone building AI tools, I wanted to go beyond just Western music and question how we could digitally preserve and reimagine traditional sounds. This project represents my curiosity, my collaboration with experts, and my belief in technology as a cultural bridge.

Research & Collaboration

Expert Interviews:Over the course of this project, I conducted interviews with three domain experts:

• A Gugak arranger & pianist (Taesoo)

• An MIR researcher from KAIST (Danbi)

• A singer-songwriter and independent producer in NYC (Joowon)

These interviews shaped my understanding of the technical, aesthetic, and creative gaps in current AI music tools—particularly for genres like Gugak that rely heavily on improvisation, microtonality, and dynamic ornamentation.

Key Insights:

• AI is useful as a co-creator, not a replacement.

• MIDI systems fail to capture the essence of Gugak due to its microtonal scales and rhythmic flexibility.

• Sampling-based tools and pattern-based composition may offer a more culturally sensitive approach.

A Gugak arranger & pianist (Taesoo)

An MIR researcher from KAIST (Danbi)

A singer-songwriter and independent producer in NYC (Joowon)

Prototyping & Technical Implementation (for Visual Design & Craft)

Challenge:How do we turn Jeongganbo (정간보)—a vertically read Korean notation—into something AI and musicians can use?

My Approach:I built a prototype web converter that:

1. Accepts Jeongganbo as a .txt file

2. Translates each symbol into Western notation using music21

3. Outputs MIDI and PDF sheet music (via MuseScore)

Tech Stack:

• Python (Flask for web interface)

• music21 for symbolic music parsing

• MuseScore for MusicXML to PDF conversion

• ngrok for temporary web hosting

💡 Design Decisions & Reflection (for Narrative & Framing)

From Exploration to PurposeThis wasn't just about building a converter—it was about creating something respectful. I intentionally avoided fully automated music generation and instead focused on tools that allow interaction.For example, the converter only gives a base melody; users are encouraged to adjust rhythm, ornamentation, and instrumentation.

What I Learned:

• AI needs cultural context to be truly useful.

• Collaboration with human musicians is essential.

• Over-automation can erase what makes traditional music beautiful.

Future Vision (for Signature & Voice)

What’s Next?

• Develop a Gugak MIDI plugin that lets composers drag-and-drop rhythmic loops and ornaments

• Integrate AI-guided recommendations (e.g., suggest Sigimsae for melodic variation)

• Partner with Gugak institutions to build open-source datasets for research and tool development

Conclusion

By treating AI not as a composer, but as a collaborator, we can preserve the soul of traditional music while expanding its reach. Gugak doesn’t need to be “Westernized” to go global—it just needs tools that understand its logic and respect its nuance.