Yalin Wang · CLIENT: Yalin Wang
Algorithmic Flesh is an interactive installation created for the presentation of cultural heritage. The installation tracks the real-time body movements of a participant and uses the movement data as the input for a generative artificial intelligence system. The system is trained on images of Dunhuang murals and generates a celestial dancer in motion on a large projection. When a posture matches a recorded mural pose, the participant's silhouette is transformed into the dancer. The aim is to turn a static archive into a shared and living experience.
Press Kit
- Unique Properties
- Algorithmic Flesh is an interactive installation redefining cultural heritage preservation as Generative Revitalization. It introduces Somatic Prompting: bypassing text interfaces, the system uses the participant's real-time physical movements as the generative seed. This somatic input drives a deeply trained AI to computationally hallucinate a dynamic, multi-limbed celestial dancer in real-time, using data from UNESCO Dunhuang murals. The audience becomes active co-creators of a living ritual.
- Dimensions & Packaging
- As a digital installation, "Algorithmic Flesh" is highly scalable for gallery or museum contexts. The typical setup requires a projection area of at least 3m x 2m (supporting either 16:9 landscape or 9:16 portrait orientation, up to 4K resolution) and a participant interaction zone 3-4 meters from the screen.
- Tags
- Generative AI, Embodied Interaction, Digital Ritual, Cultural Heritage, Human-AI Co-Interpretation, Somatic Prompting, Interactive Art Installation, Dunhuang Murals, Algorithmic Art, Cultural Computing, Choreography
- Production Technology
- Algorithmic Flesh uses a custom, closed-loop machine learning pipeline within ComfyUI. Real-time, markerless biomechanical tracking is achieved via OpenCV, MediaPipe, and YOLO26. Visual synthesis is driven by a FLUX LoRA model, fine-tuned on Tang Dynasty Dunhuang murals for absolute aesthetic fidelity. Temporal coherence is managed by Wan 2.2 Mixture-of-Experts video architecture, while the Google Veo 3 API provides cinematic spatiotemporal synthesis and native, physics-based audio generation.
- Design Challenge
- The primary challenge was epistemological: How to reconstruct the dynamic, anti-gravitational flight of ancient murals in a virtual space without reducing them to lifeless 2D data? We emphasize not replacing the physical UNESCO Mogao Caves with a simulation, but speculating on new representation models. We challenge the passive museum dynamic, proposing an interactive framework where human flesh computationally hallucinates and sustains the continuous flow of a broken historical lineage.
- Project Duration
- The project was developed over a 10-month period from January to October 2025. The research and technical implementation were conducted in Beijing, China, in affiliation with the Central Academy of Fine Arts.
- Operation Flow
- A participant stands before the projection. Instantly, a "Cosmic Telemetry" UI tracks their skeletal data in real-time. As they move, the system calculates their posture against a canonical Dunhuang dataset (e.g., "raised_leg_flying [0.96] Mogao Cave 285"). Upon a high-confidence match, this somatic prompt triggers the AI. The UI dissolves, and the user's silhouette morphs into a fluid, hyper-realistic celestial dancer, transforming the viewer into a co-creator of a living ritual.
- Research
- This research addresses the loss of kinesthetic knowledge in static cultural heritage by introducing a Generative Revitalization framework. Pioneering somatic prompting, the project explores how a participant's body guides a generative AI without text. The machine, trained on Dunhuang mural archives, computationally generates lost transitional movements between ancient poses. This offers a new perspective on digital materiality, transforming a passive archive into an embodied, co-interpretive ritual.
- Inspiration
- How can we reawaken a dance frozen in stone for a millennium? Traditional digital preservation secures visual fidelity but flat-lines the dynamic essence of the Dunhuang Apsara murals into static maps—a Crisis of Stasis. Inspired by Heimkehren (returning home) and Buddhist philosophies of Anitya (Impermanence) and Pratityasamutpada (Interdependence), the design treats machine intelligence as a collaborator, celebrating the poetic instability of forms in transition as Digital Epiphanies.
- Image Credits
- Image #1: Creator Yalin Wang, Algorithmic Flesh, 2025.
Image #2: Creator Yalin Wang, Algorithmic Flesh, 2025.
Image #3: Creator Yalin Wang, Algorithmic Flesh, 2025.
Image #4: Creator Yalin Wang, Algorithmic Flesh, 2025.
Image #5: Creator Yalin Wang, Algorithmic Flesh, 2025.
Video Credits: Creator Yalin Wang, Algorithmic Flesh, 2025. - Patents
- Copyrights belong to Yalin Wang, 2025. All algorithms and visual outputs are original research generated for academic and artistic purposes at CAFA.
Press releases are available in 1 languages.