Gallery

Jason Bailer Losh

Different denotes neither bad nor good, but it certainly means not the same

Jason trained custom models to output images used in the composition of a new large scale oil painting. He trained two LoRAs, one on his own sculptures and another on an archive of photographs he's amassed of living rooms from the 60s, 70s, and 80s.

September 27, 2024

Drew Nikonowicz

Photographic Diagnostics

Beginning with a single 4x5 film photograph made in early 2024, Drew's goal was to use LoRA training in "incorrect" ways, to attempt to push its boundaries and make it break down. He was interested in using AI tools in "diagnostic" ways, to reveal more about his own practice and potentially reflect on the implications of these new tools themselves.

ZERO is a psychological sci-fi short film in the tradition of Solaris, Lost, Annihilation, and many others. During his residency, Noah developed a full script, 22 minute animatic, and completed the Station Zero Orientation video, which plays on-station as a diegetic intro video early in the film. The goal is to create a 20 minute animated film utilizing live actors in concert with AI models to create vid2vid animations that replicate a traditional animation pipeline but allow for the fidelity of emotional human performances. As of July, 2024, ZERO is still in production with ongoing support from Late Night Labs.

A collection of vignettes centered around various odd birds in odd royal situations, woven together into a short film. After a lot of trial and error, I finally created a 'stop motion AI workflow' with a lot of human control/input and that works well with the environment, quick movements, and multiple characters. The process is a mix of everything I learned (and some new stuff) in the process of creating the odd birds project.

Domenec Miralles

Infinite Mesh is both a story generator and an ongoing meditation on automation. From a given prompt, this program outputs a video file and an accompanying comic book. At its heart lies the question: how can a synthesis between large language models and stable diffusion image models create a semblance of originality?