“Feeling the Flow”: Evaluating a Hybrid AI-Fluid Simulation System for Embodied Artistic Expression

Source of Publication

Proceedings of the 31st International Symposium on Electronic/Emerging Art (ISEA2026)

Conference Year

2026

Abstract

Real-time AI image generation is increasingly used in live performance, but often behaves like an autonomous visual engine: small changes in movement do not lead to clear, controllable changes in the output. This paper presents a hybrid system that inserts an artist-controllable fluid simulation between performer gesture and a real-time diffusion model, aiming to narrow this control–expression gap. Performer movements drive a GPU-based fluid field whose evolving patterns are used as dynamic prompts for image generation. In an expert study with seven practitioners from theatre, dance and new media art (N = 7), we compare this hybrid configuration with a pure camera-to-AI baseline. Experts reported that the baseline felt unpredictable and hard to “drive”, while the fluid layer made their influence on the visuals more visible and negotiable, making the system more suitable for choreographed performance while still allowing generative surprise. At the same time, systemic issues in current real-time diffusion—visual flicker, identity drift and high computational load—remained clear constraints. We contribute this empirical comparison plus design observations on fluid simulation as an embodied, physically intuitive mediator for AI-driven performance visuals.

Document Type

Conference Proceeding

Publication Date

11-6-2026

This document is currently not available here.

Share

COinS