Danilo Lieder holds a BA in Philosophy from the University of Vienna and is currently an MA (CE) student in Game Studies at the University for Continuing Education Krems. His work in game studies focuses on game design, the philosophy of play and games, phenomenology, and aesthetics. He is also the founder and game designer of the independent game studio LudicOwl, currently being established.
Prompting the Belle Époque: Building a Playable Art History with Generative Assets
FROG 2026 – Talk
Répoque, a solo RPG Maker MZ project in early production, is a coming-of-age RPG set in a fictional Russian-Swedish-inspired spa town around 1900 – a walkable memory rather than a historical reconstruction. Its world is assembled from art-historical styles that operate as playable semantic registers: aged oil realism carries the remembered everyday; rococo gold varnish marks the spa quarter and its class boundaries; ink illustration renders folklore and documents; Art Nouveau belongs to the children’s imagination, where playground battles become pulp-novel sieges. The same scene, shifted through a register, keeps its diegetic content and changes its meaning.
Generative image models turn such registers into controllable production variables. My design work has shifted accordingly – less image-making, more “register curation”: selecting corpora, building prompt scaffolds, enforcing consistency rules, documenting interventions. This, I argue, amounts to a museum analogy in practice: what a museum does with its collection – selecting, framing, recontextualizing – an AI-assisted asset pipeline does with the art-historical corpus it draws on.
The talk presents this practice as a case study – one town, four registers, one pipeline: a restorative workflow that keeps every asset’s geometry from the original, restricts generation to material and style, and unifies the result under a deterministic “varnish” of darkened memory. A transposition method translates 1990s childhood functions into 1900 equivalents – deliberate, disclosed anachronism, set against the anachronism noise the models produce unasked. The project is part of ongoing master’s research on emerging technologies in computer game asset production.
Yet the practice sharpens a question rather than settling it: what does “period-appropriate” mean when the period comes out of a diffusion model? Learned styles arrive without attributable provenance – style as promptable commodity; and a statistical average of stylistic markers is not a work in a style – the kitsch risk peaks precisely where ornament is densest, in Art Nouveau. I propose treating the pipeline as a curatorial instrument rather than an automaton – and “synthetic heritage” as a shared problem that museums and game production may be better off negotiating together.