Latent Imaging and Imagining

Trans(-feminine) temporality

This work is rooted in an instance of trans(-feminine) temporality, in which transness and processes of transitioning have enabled and positioned the artist outside a chrononormative biography and life schedule, where from a normative viewpoint childhood, adolescence and adulthood stand in temporally discontinuous relation towards each other: A friction that enables one to explore modes of becoming an embodiment of alternative spatiality and temporality, not determined by the constraints and implications of cis-heteronormativity and the gender binary. Besides the resulting joy, euphoria and pleasure though, this friction also causes certain stresses and dysphoria, such as the felt “loss” or inability of having been experiencing girlhood. Latent Imaging and Imagining as a methodology therefore attempts to care for this situation by layering generated imaginative imagery as latent versions of the artist's childhood photos and memories, thus balancing and negotiating degrees of maintenance, disruption and repair.

Latency

Latency in the context of this work references not only a delay, but also the “virtual”, in the meaning of existing or “present but needing particular conditions to become active, obvious, or completely developed” [1]. A latent imaging and imagining therefore describes the practice of visualizing an imaginary that finds itself in the realm of the virtual and contingent - possible to become factual, but not (yet) unfolded.

Passing as "girl" to an AI

An important critical moment in this work can be found in how the imaginary images take their form: Based on a dataset of millions of images labelled as "girl" [2], the used AI tool generates images which the AI considers as female-passing. So not only does the AI come from a stochastic process of coming to terms with what it classifies as "girl", what might also be called a potential bias, considering it is not transparent what images specifically have been used. Also, the idea of what passes as "girl" (used as a prompt) to the AI, can be observed in the resulting images: A cute white blonde healthy middle-class-looking girl. Only adding more detailed descriptions to the prompts, such as "girl with brown hair", "a young girl in a t-shirt", led to better, more realistic results for the purpose of this work.

Lastly should be noted, one might think that the generated images, depicting a white girl by using only the prompt "girl", are resulting from the context of the input images, which actually contain the face of a white child. This could be true and probably is the reason in this case. Nevertheless, a quick experiment [3] of generating 40 images using only the prompt "girl" shows that this prompt mostly defaults to depictions of light-skinned girls, and more diverse image results are still an exception with the used AI tool.