Horomata
From the Greek ὁράματα: visions, sights granted in a dream.
Horomata is the name I gave to the body of generative work I began in 2021, before generative AI became a consumer tool, when getting an image meant building and training your own models: programs written and adapted in Python, datasets assembled by hand, corrected epoch after epoch. No product existed yet.
I trained networks on the complete filmographies of directors with strong visual signatures, Ozu, Sirk, Lynch, Kubrick, then pushed the work toward text to image, building a common language with the machine: stopping it when it drifted, redirecting it toward the theme, the structure and the aesthetic I wanted, letting it surprise me when the surprise served the piece. That posture, directing the machine rather than operating it, is the same one I work with today. Only the tools have caught up.
The work was published in a book in June 2022 and publicly timestamped online in January 2022. The pieces below are shown as they were made, with the grain of their time.
Horomata, the book. Published June 2022.
Childhood at war. 2021. Model trained in Python.
Burning News from Exile. 2021. The spectacle of misfortune, and what it erases.
Torso series. 2021. Two states of one investigation.
Hibagon ヒバゴン. Series of 8, 2021.
The Hibagon is Japan's Bigfoot, a creature said to roam Mount Hiba in the Hiroshima Prefecture. Sightings were so frequent in the late 1970s and early 1980s that the county opened a dedicated office and Kobe University ran investigations. I worked only from the eyewitness reports, using their descriptions to build photorealistic portraits of a creature no camera has ever caught.
Yamabiko 山彦. Series of 12, 2021.
Yamabiko, literally "echo", is a mountain yōkai of Japanese folklore, the spirit behind the sound that answers back from a canyon. Described as a small creature between a dog and a monkey, it lives deep in the mountains, often heard and rarely seen. Continuing the Hibagon work, I generated these twelve images from descriptions scattered across different texts, letting the creatures blend further into their surroundings until the vegetation and their coats begin to merge.
Ashes to Ashes. 2021. Generative video, made when this was pioneering work: no product existed, every step built and trained by hand. An AI face first modeled on the Louvre Aphrodite, then aged and rejuvenated by AI into a full lifetime, then turned into a video where the portraits morph into one another in a single seamless movement from birth to death. Shown here in an accelerated fifteen minute cut of the original.
Horomata is the root of everything on this site. The eye came first; the tools caught up.
← Back to current work