Human-AI Co-Dancing: Evolving Cultural Heritage through Collaborative Choreography with Generative Virtual Characters

Abstract

This research introduces an approach for translating traditional dance knowledge into interactive computational models extending beyond static dance performance recordings. Specifically, this paper presents the concept of "Human-AI co-dancing," which involves integrating human dancers with virtual dance partners powered by models derived from dance principles. To demonstrate this concept, the research focuses on the choreographic principles deconstructed from the knowledge of traditional Thai dance. The principles are analyzed and translated into computational procedures that dynamically manipulate the movements of a virtual character by altering animation keyframes and the motions of individual joints in real-time. We developed an interactive system that enables dancers to improvise alongside the virtual agent. The system incorporates voice control functionality, allowing the dancer, choreographer, and even the audience to participate in altering the choreography of the virtual agents by adjusting parameters that represent traditional Thai dance elements. Human-AI rehearsals yielded intriguing artistic results, with hybrid movement aesthetics emerging from the synergy and friction between humans and machines. The resulting dance production, "Cyber Subin," demonstrates the potential of combining intangible cultural heritage, intelligent technology, and posthuman choreography to expand artistic expression and preserve traditional wisdom in a contemporary context.