New Work

Welcome to my newsletter era ♥. My latest piece, "VŌX, 2024" was just exhibited in Tokyo. Check out unreleased b-sides and get a sneak peak into my tech stack.

Aug 16, 2024

VŌX, 2024 — photography, novel AI checkpoint, diffusion output. Exhibited at NOX Gallery in Tokyo, July, 2024. Collect it on Zora

On assignment for Feu Magazine, I traveled to Porto, Portugal to meet with musician and artist Vōx. We spoke about her forthcoming album, neurodivergence, childhood trauma, teenage angst and technological escapism, some color and context for the album that I had received a preview of. After chatting, we dove into wardrobe and began the photoshoot. As a long time photographer, it was an interesting mental exercise to make images that I knew wouldn’t be seen as the final result, but would instead be used to train a new AI model that I could mix and mold to create an infinite amount of results.

Back home in Lisbon, I compiled 30 images from the shoot that I felt captured her essence and would allow me to explore her visage in a new way. I make my AI models the old fashioned way, creating entirely new checkpoints from scratch using Fast Dreambooth, a clever bit of free, open source software. Dreambooth intakes the images, a new key phrase, and trains against the existing SD 1.5 model. (Now considered obsolete given the emergence of Flux and SDXL, oh well - I like it!)

the training data

1.6 gb checkpoint in hand, I jumped over to Deforum’s open source notebook and plugged the model into the software like swapping in a game cartridge. From there, I began exploring dozens of aesthetic options, dialing in prompts and settings until it was consistently putting out images I felt encapsulated the conversation I had with Vōx during the shoot.