Clip+Replace, 2024

Stable Diffusion, custom algorithm, self-portraits

Click to load imageAI-transformation of original self-portrait from mid 2023.

To reduce website CO2e emissions, project images need to be manually loaded.

AI-transformation of original self-portrait from mid 2023.

Click to load imageClose-up of AI-transformation of original self-portrait from mid 2023.

To reduce website CO2e emissions, project images need to be manually loaded.

Close-up of AI-transformation of original self-portrait from mid 2023.

Click to load imageClose-up of AI-transformation of original self-portrait from mid 2023.

To reduce website CO2e emissions, project images need to be manually loaded.

Close-up of AI-transformation of original self-portrait from mid 2023.

Click to load imageClose-up of AI-transformation of original self-portrait from mid 2023.

To reduce website CO2e emissions, project images need to be manually loaded.

Close-up of AI-transformation of original self-portrait from mid 2023.

Click to load imageAI-transformation of original self-portrait from mid 2024.

To reduce website CO2e emissions, project images need to be manually loaded.

AI-transformation of original self-portrait from mid 2024.

Click to load imageClose-up of AI-transformation of original self-portrait from mid 2024.

To reduce website CO2e emissions, project images need to be manually loaded.

Close-up of AI-transformation of original self-portrait from mid 2024.

About

Clip+Replace is a series of self-portraits transformed by a confused AI algorithm.

The series takes a critical glance at AI image generation technology, and shows how it can create novel aesthetic expressions that simultaneously reveal how the technology (mis)represents reality. At first glance, the works in Clip+Replace appear to be coherent self-portraits, but a closer look quickly reveals that each photograph has instead been replaced by multiple smaller images with bizarre subject matter. Through a custom built algorithmic process, hundreds of parts of the original image have been replaced by AI-generated versions that look somewhat similar to the original, but not quite. In these new smaller images, we find eyes that are multiplied, pink hair which becomes a woman’s face, cats that appear out of thin air and many other eerie occurrences. Upon seeing these surreal juxtapositions, one is left wondering what is real and what is AI.

The title of the series is a combination of copy+paste and CLIP, which is the image classification technology that lies at the heart of all image generation AI. The title also describes exactly what the custom algorithm behind the series actually does. It uses CLIP to describe a part of the image, and then it replaces that part with a newly AI-generated image based on the CLIP description. In doing so, the algorithm tries to leave as much as possible of the transformation process in the hands of the AI systems. The result is something that seems real at first, but is soon revealed to be a strange mess of incoherent data points, much like most AI-generated content.

Both the algorithm and the original self portraits are created by myself. The original photographs are rejected images from my ongoing performative photo project “in transitu” in which I document my gender transition on a weekly basis. In Clip+Replace, these rejected originals are then subjected to my own algorithm, which uses two AI systems in tandem to create these odd collage-like images. The AI systems reveal some of their own biases when they gender my body in various ways by feminizing or masculinizing my lips, breasts, eyelashes and more. They also imagine things in the white curtains that are not there in the originals, such as cats, birds and airplanes. The transformed images thereby end up bringing to light the distorted reality of AI image technologies, while at the same time demonstrating an aesthetic unique to AI-based algorithmic processes.

CO2e used: ~0.00g