A generative art practice exploring what image models see when they look at history. Currently producing Exposure-series — a 144-piece portrait matrix.
Vanguartisan is a creative practice built around a single question: what can generative image models reveal that conventional image-making can’t?
Not illustration. Not replacement. Something more specific that uses the systematic properties of these tools (the way they weight language, the way they interpolate between training references, the way they fail) as material in themselves.
The first major body of work is Exposure-series: a 144-piece generative portrait matrix.
Twelve historical figures. Twelve twentieth-century portrait photographers. Each figure photographed by each maker — a 12×12 grid where every piece exists in two readable directions at once.
The premise is a specific anachronism: portrait photography arrived too late for any of these figures. The medium that developed the tools for catching power off-guard — Karsh, Avedon, Arbus, Halsman — never had access to Caesar, or Wu Zetian, or Joan of Arc. Exposure-series is an attempt to close that gap, not nostalgically but structurally. No grain simulation. No daguerreotype aesthetic. Clean, contemporary photographs. The anachronism is the point.
Running something this systematic required building a pipeline rather than a prompt list. Three agents, each with a defined scope:
The Series Director holds the conceptual brief — the premise, the edition standards, the failure modes that disqualify an output from the series.
The Figure Researcher does primary source research before any generation begins. Ancient texts, surviving portrait artefacts, numismatic evidence, modern scholarly reconstruction. Nothing goes to the Promptsmith without an approved physical brief.
The Promptsmith translates the approved brief into generation-ready prompts. It knows the technical syntax of each tool, the guidance parameter ranges, how Flux2 weights word order.
The pipeline exists to prevent the model’s training data from determining the figure’s appearance. Without it, you get marble busts — the posthumous idealisation, because that’s what the training data is saturated with.
Caesar-series is the first sub-series through the full pipeline. Reference images are being established. The full 12×12 matrix is the complete work.