Exercises in Style

The Original Game

In 1947, Raymond Queneau posed a radical question: is it possible to submit a base text to all conceivable variations, provided each follows some rule? His answer was Exercises in Style ("Exercices de Style") — 99 retellings of the same trivial anecdote, a minor disagreement on a bus, each transformed through a different narrative lens: from haiku to mathematical notation, from official letter to dream sequence.

As Umberto Eco observed in his notes on the Italian translation:

"The comic effect is global, born from accumulation [...] while we laugh at a mechanical exchange of alphabetic letters, we laugh simultaneously at the author's wager, at the balancing acts he employs to win it, and at the nature of both a given language and the faculty of language as a whole."

Read Umberto Eco's complete notes in the original PDF.

A New Match

Eco wrote that fidelity to Queneau meant "understanding the rules of the game, respecting them, and then playing a new match with the same number of moves." This project is exactly that: a new match of Queneau's game, played with artificial intelligence.

The same question Queneau asked of language in 1947 is now posed to AI: can the same story generate infinite visual worlds simply by changing the words that describe it? Using text-to-video models, each variation reveals how AI interprets tone, style, and linguistic nuance — transforming identical narratives into radically different visual experiences.

Words as Creative Tools

Each video begins with the same core narrative but employs different vocabulary, sentence structures, and stylistic approaches. The results are remarkably diverse — proving that in the age of generative AI, words are not mere instructions but creative tools that shape the entire aesthetic and emotional quality of the output. The accumulation of variations, as Eco noted, creates meaning through multiplicity itself.

Co-Created: The Artist in the Age of Intelligent Machines

Exercises in Style was exhibited at the BCA Center in Burlington, Vermont as part of "Co-Created: The Artist in the Age of Intelligent Machines", a group exhibition curated by Chris Thompson. The show ran from February 10 to May 6, 2023.

The exhibition presented eight artists who use machine learning as an artistic medium. As stated in the exhibition materials: "Probing the creative and technological boundaries of learning machines, these artists have been hacking systems, gathering data, and training neural networks" - work that predates the emergence of popular generative AI tools.

Alongside my work, the exhibition featured pieces by Jane Adams, Memo Akten, Minne Atairu, Lapo Frati, Jenn Karson, Casey Reas, and Jason Rohrer.