Conjured Terrain

About

Vellum LA is pleased to introduce Conjured Terrain by Casey REAS, curated by Jesse Damiani, Alice Scope, and Sinziana Velicescu. The exhibition will take place from October 7 – October 29, 2023. The exhibition is framed as a call-and-response with DAM Projects in Berlin. All work will be presented at Vellum LA on state-of-the-art Luma Canvas LED displays.

Casey REAS, the “Godfather of Generative Art,” is a pillar of the Los Angeles art community, and one of the most important and beloved new media artists working today. Conjured Terrain presents all-new work from the artist’s acclaimed, ongoing Untitled Film Stills (UFS) project, as well as associated videos from the related body of work, Compressed Cinema.

REAS’ work stands in contrast to popular narratives about art created using code, especially given the recent hype around AI image generators, which have been trained on vast datasets and are able to produce plausible imagery in many styles when a user “prompts” them. With Untitled Film Stills, REAS engages machine learning as material not to quickly produce a functional image, but instead to thoughtfully examine imagery’s ability to contain and convey realities.

“Each Untitled Film Still image is a frame from an imagined film,” REAS says. “The idea is to reveal another reality within the original film, something that has the character of the original work, but is transformed into something else. A film is a perspective from which to view the world, and a set of Untitled Film Still images are a unique view of the original film that produces a new and alternate reality in an unexpected way.”

The process begins with the artist selecting a film, which is converted into hundreds of thousands of individual frames (24 frames for each second of the original media). A machine learning model called a generative adversarial network (GAN) is then trained on those frames, by which it learns to generate imagery in the style of the original films. This process—like shooting some films—takes about a month. After the model is trained, REAS generates thousands of new images from the indefinite number of possibilities contained within the model. Sometimes, a different machine learning model is used to change the visual tone of the images. These synthesized images become the artist’s “raw footage.” REAS then curates these into smaller numbers and places them in a sequence, after which they go into “post-production,” where REAS resizes and color grades them.