Hunziker

Description

Screen Test #32 is part of a series of AI-generated portraits I created for a video project of the same title. Each portrait features a person posing with an unfamiliar organic presence — not as spectacle, but as quiet, domestic intimacy. The grayscale palette of these fictional portraits creates a sense of timelessness, suggesting photographic realism reminiscent of an archival ethnographic study, a contemporary fashion portrait, or a frame from an untold science fiction film. The individuals in these portraits show traces of time, wear, and experience — resisting the polished standards of algorithmic beauty. Their lived, unlived lives suggest a new kind of mythology: not one of future technological triumph, but one marked by domestic strangeness, hybrid identities, and unresolved complexity. Trapped in a loop of past data, without clear boundaries between object and subject, AI portraiture becomes a kind of ethnographic practice — not of distant cultures, but of the mirrored strangeness within ourselves.

Process

Screen Test #32 emerged from a desire to challenge the normative aesthetics produced by generative AI systems — particularly the recurring erasure of age, complexity, and affect in female representation. I was interested in creating a portrait of a woman marked by time, by narrative, by the texture of a life lived or imagined — a figure that resists the algorithmic default of youth, beauty, and seduction. As an artist working critically with AI, I have become increasingly aware of how deeply these tools are shaped by the training data they ingest. These datasets are not neutral; they reflect existing cultural biases, especially around gender, beauty, and desirability. Moreover, commercial AI systems are often layered with safety filters that suppress any output considered unsettling, ambiguous, or emotionally charged — reinforcing a sanitized visual language. The result is a homogenous visual culture: smooth, upbeat, emotionally flattened. In my practice, I seek to interrupt this visual consensus. I approach AI not as a tool for generating idealized images, but as a medium to expose the tensions between the synthetic and the human. Screen Test #32 is part of this inquiry: a portrait that refuses perfection, instead offering a speculative subjectivity — one shaped by the aesthetic residue of a fictional life. It is both artifact and counter-image, proposing an alternative visual mythology within the closed loops of machine learning.

Tools

I started with a basic prompt in Stable Diffusion 1.4, creating several rough sketches, before continuing in Stable Diffusion XL 1.0 and using some of these drafts as reference images. I did some basic fine tuning and cleanup in Photoshop and then upscaled the image in Topaz Gigapixel.