Algorithmic Flesh Yalin Wang, 2025
The Dunhuang Apsara murals, a UNESCO World Heritage site, represent a dormant interactive interface whose original kinesthetic and ritual function is lost. Current digital heritage approaches often perpetuate this gap by focusing on static preservation. We present Algorithmic Flesh, an interactive installation that reframes this challenge through a "Ritualistic Revitalization" framework. The system employs a real-time pose estimation pipeline to translate a participant's movements—a form of "somatic prompting"—into generative murals. Crucially, the system's AI, trained on the original artworks, generates ephemeral volumetric forms—a synthetic "algorithmic flesh"—in symbiotic dialogue with the user's body. This process makes tangible the Buddhist principles of impermanence and interdependence, echoing the murals' original contemplative function. As a contribution to cultural computing, Algorithmic Flesh offers a non-Western perspective on digital heritage. It demonstrates a new methodology for intergenerational transmission through embodied human-AI interaction, shifting the goal from preservation to a living, co-interpretive ritual that fosters personal meaning-making.







Algorithmic Flesh: An Embodied AI Co-Interpretive Ritual for Revitalizing Dunhuang Apsara Murals

This paper presents Algorithmic Flesh, an interactive multimedia installation that resolves the 'Crisis of Stasis' surrounding the UNESCO World Heritage Dunhuang Apsara murals by moving beyond conventional static digital preservation. Rather than relying on conventional text or image prompts, the system employs an alternative interaction paradigm termed "Somatic Prompting." Building upon an empirical Dunhuang dance vocabulary—reconstructed through decades of dancers' somatic practice and biomechanical analysis of the murals—the project utilizes "Speculative Fabulation" to map the participant's real-time embodied kinetics against an archive of ancient postures. Driven by a zero-latency computer vision pipeline utilizing a customized YOLO 26 model and MediaPipe, the system employs a custom ComfyUI orchestration workflow to establish a precise topological mapping between the contemporary body and archival poses. Crucially, the system's generated Apsara movement is not a mechanical restoration of historical facts, but rather a "post-human choreography." It is an inferential reproduction probabilistically "hallucinated" by the algorithm under the intervention of contemporary biological flesh and ancient aesthetic laws. By establishing a state of "Dharmic Sync," the installation unleashes creative possibilities that bridge the corporeal and the digital; specifically, it utilizes a "Nine-Grade" evaluation logic where specific poses and hand gestures (mudras) trigger the real-time mapping and generation of philosophical texts from ancient Dunhuang manuscripts. Algorithmic Flesh proposes a co-interpretive framework where ancient cultural heritage is re-enacted and evolved through the somatic agency of the contemporary viewer within a non-Western cultural computing framework.






















Copyright © 2019-2025 Yalin Wang. All Rights Reserved.