Nithya Shikarpur, MIT PhD student, gives talk on Generative Modelling and Interactive Performance for Hindustani Music

Author:

TUE 21 OCT 2025, 10:30AM, KCL WATERLOO JCMB 5.15

Nithya Shikarpur, PhD student at MIT to give talk at KCL Waterloo campus as part of the MARC Seminar Series, refreshments courtesy of the NMES Research Culture Fund.

If you are unable to attend in person, you may use the following MS Teams link to attend the event virtually : MARC Seminar Talk Nithya Shikarpur

Title : Towards Generative Modelling and Interactive Performance for Hindustani Music and Beyond

Abstract : Recent advances in generative music modelling open up rich avenues for creative exploration and expression. As both a musician and researcher, my work focuses on two interconnected goals: (1) developing generative models that meaningfully engage with the musical context and aesthetics of specific traditions, and (2) designing interactive systems that foster creative collaboration between humans and generative models. First, I will introduce GaMaDHaNi, a hierarchical generative model for Hindustani vocal music.  Through a hierarchical system, modelling pitch contours first followed by spectrograms, GaMaDHaNi captures the microtonal nuances that characterize this tradition. I will further discuss the challenges of conditioning such models on musically relevant parameters such as raga. Second, I will share insights from β€œcat-in-loop”, a collaborative performance created with the Cat in Black ensemble and my collaborators Weilu Ge and Hugo Garcia.

This work explores the creative (mis)use of VampNet, a masked audio transformer model, as a tool for embodied improvisation and human-AI co-performance. Together, these projects reflect a broader inquiry into how generative systems can engage with musical practices – not merely as data, but as living, evolving traditions of sound, gesture, and interaction.

Speaker : Nithya Shikarpur is a second-year PhD student at MIT supervised by Prof. Cheng-Zhi Anna Huang. She is interested in the modeling of human-AI interactive systems for music creation and creativity especially for low-resource genres of music. Earlier she was at UniversitΓ© de MontrΓ©al and Mila for her MSc also with Prof. Huang. Nithya is also a practitioner and performer of Hindustani classical vocal music and draws on this knowledge to further her research projects.