Nils Gallist is a researcher and media artist focusing on interactive media, games, and immersive experiences. His work combines artistic practice, technological experimentation and research into playful forms of interaction.
Juergen Hagler is an academic researcher, artist, and curator specializing in animation, media art, and interactive art. He is Professor of Animation and Media Studies and Head of Digital Arts at the University of Applied Sciences Upper Austria. His work bridges artistic research, emerging technologies, and curatorial practice.
Best Practices for Real-Time Generative AI in Participatory and Playful Cultural Installations
FROG 2026 – Talk
Real-time generative AI increasingly enables audiences to step from passive viewing into active, playable participation in cultural and artistic content. Yet moving image-generation pipelines from controlled demos into unattended public installations, which are running for hours daily over weeks or months, surfaces practical and conceptual challenges rarely discussed in technical literature. This paper distills best practices and failure lessons from two deployed real-time AI installations built by our team: The Theater Machine (Landestheater Linz, 2025), which transforms foyer visitors into characters from canonical stage productions via a StreamDiffusion pipeline, and The Fadinger Automat (Schlossmuseum Linz, 2026), which continuously reimagines visitors as historically inspired portraits of a figure with no surviving authentic likeness over a six-month exhibition run.
Drawing on these cases, we discuss recurring technical and design tensions: balancing latency and visual fidelity for a “playable” real-time feel; maintaining consistent character or stylistic identity across a continuously drifting diffusion stream; designing low-barrier, intuitive interaction for unsupervised public audiences with no technical background; and handling the ethical weight of AI-generated representations of real people, historical figures, and cultural heritage. We also address production-side lessons such as: hardware reliability for unattended multi-month operation, content moderation for open-ended visitor input, and sustainability considerations across ecological, social, cultural, and economic dimensions.
Rather than treating these installations as isolated art pieces, we frame them as case studies in “playing art”: using generative AI to invite audiences into an active, game-like relationship with it.