On May 30, Mihir Raina, performing under his artistic alias Lacuna, presented the results of his Open Culture Tech residency during a live showcase at De Helling in Utrecht. The performance introduced audiences to Saz-e-yaad, a project exploring how AI can help preserve, reinterpret, and extend cultural memory through music. Developed in collaboration with AIxDESIGN, Saz-e-yaad combines genAI, archival recordings, and immersive audiovisual storytelling. Drawing from rare recordings of Kashmiri folk and classical music, the project bridges cultural heritage with contemporary electronic music, creating new ways of engaging with tradition through technology.

 

 

 

 

Reimagining heritage through sound

 

Originally from India and now based in Utrecht, Mihir Raina's artistic practice centers on the reinterpretation of heritage sounds through electronic music, sound system culture, and emerging technologies. His work explores how archival materials can be transformed into new sonic experiences.

 

As Lacuna, Raina blends electronic sound design with Kashmiri musical traditions that have been shaped by migration, political divisions, and oral transmission. His research investigates how digital audio workstations, hardware instruments, and AI can be used to creatively process and contextualize heritage audio. At the heart of this work are questions surrounding cultural memory, migration, sound archivism, and decolonial approaches to technology. 

 

 

Introducing Saz-e-yaad

 

During the residency, Raina and the AIxDESIGN team collaborated with Moisés Horta Valenzuela (Hexorcismos) to develop a custom standalone app trained on recordings from Ravimech Studios, including rare Kashmiri folk and classical music from the 1980s and 1990s. The software allows users to generate new audio inspired by the source material through a simple interface. Users can choose between models focused on different musical elements, including melody, harmony, rhythm, or complete compositions, and guide the generation process through text prompts, audio prompts, or fully autonomous generation.

 

Rather than recreating existing recordings, the instrument generates entirely new musical material rooted in the characteristics of the original archive. The result is a workflow that enables forgotten rhythms, vocal textures, and musical ideas to be revisited in contemporary forms.

 

 

Performing live in Utrecht

 

The residency concluded with a live performance at De Helling in Utrecht, which immersed audiences in a powerful blend of bass-heavy electronic music, experimental sound design, and audiovisual storytelling. Using outputs generated with Saz-e-yaad alongside live performance techniques, Lacuna created a set that moved between intensity and introspection. Layers of rhythm, noise, and melody merged into an experience that felt both futuristic and deeply connected to tradition.

 

Audience members described the performance as mysterious, innovative, and highly immersive. Many were surprised by how naturally the AI-generated material blended with traditional musical elements. Others noted the careful use of tension throughout the performance and the way it balanced experimentation with emotional depth. Rather than focusing on the AI, audiences connected with the music, the storytelling, and the emotions it evoked.

 

 

Building the tech

 

Over the course of the residency, the team developed one standalone app and three genAI models, all trained on Kashmiri folk recordings sourced from Ravimech Studios. The software was designed to run on entry level hardware, demonstrating the accessibility of the technology.

 

The flagship model, Saz-e-yaad Ravimech, was trained on complete musical recordings and generates full-spectrum outputs that combine melody, accompaniment, rhythm, and texture. A second model, Saz-e-yaad Harmony, focused on melodic and harmonic material by training on recordings with percussion removed. The resulting outputs often produce atmospheric and dreamlike musical phrases. The third model, Saz-e-yaad Rhythm, was trained on isolated percussion recordings and proved particularly effective at capturing rhythmic patterns and maintaining musical groove.

 

Together, these models form the foundation of the Saz-e-yaad instrument, allowing users to explore different aspects of Kashmiri musical heritage through generative audio.

 

 

Looking ahead

 

What began as an Open Culture Tech residency evolved into a broader exploration of how AI can engage with cultural heritage in meaningful and creative ways. Across several technical experiments and a public showcase, Saz-e-yaad demonstrated that generative technologies can do more than reproduce existing material: they can create new possibilities for preserving and reimagining cultural memory.

 

For Mihir Raina, the project represents an ongoing investigation into how archives can become living, creative resources rather than static records of the past. By bringing together heritage sounds, electronic music, and machine learning, Saz-e-yaad offers a glimpse into how technology can help carry cultural stories into the future.

 

https://www.instagram.com/lacunafactory/ 

 

https://aixdesign.co/projects/open-culture-tech