The Interzones revisited
Technical process
I experimented in ComfyUI with the Juggernaut model as a checkpoint (Stable Diffusion), texture injections using IP adapters, upscaling processes, and ControlNet to preserve the structure, I guided the original sequence through AnimateDiff to create subtle textures: rust, cracks, imperfections... In a two-iteration process, I pushed the models to their limits to explore how far I could stretch the original video without completely breaking the mathematics of the latent field. The piece's native artifacts—small glitches and errors originally created with data moshing—were introduced into the flow purely out of curiosity to see how KSampler would interpret and assimilate this digital noise. To complete the entire process and experiment with iterations, I used compositing nodes in Blender and color curves to make subtle adjustments. The narrative circle closes with a sound layer created on a classic MicroKorg, seeking a synthesis of subterranean textures to accentuate the liminal aspect of the work.
Origin and motivation
Coinciding with the recent closure of the Foundation platform and my ongoing research into ComfyUI and locally executed AI models, I wanted to take this opportunity to relaunch “The Interzones” The original piece was born in 2024 as part of the collective drop “AVU Megavision” by the “Analog Video Union” collective, and was later exhibited at RGBTML in Montreal. That first version never sold, and in 2026, after the disappearance of the original mint due to Foundation's collapse, I found the perfect excuse to merge my research in ComfyUI with a second life at Transient Labs, adding an extra layer of imperfection and uncanniness that deliberately destroys the original smoothness of the 3D video. My primary motivation for using AI is not only aesthetic but also conceptual: to explore beyond the structuralist outcome offered by 3D and experiment with the metaphysical and semantic weight of the prompt, where the word becomes the chisel of the texture.