As with many of my projects, I establish a few constants that serve as a framework, allowing me to experiment with various directions while staying true to a singular theme. This study began with original studio photography, which was then manipulated through a neural pipeline to evolve into a series of fluid, temporal animations.

The Constants

The visual language of this study was restricted to three core elements: Lips, Spheres, and the refractive texture of Mylar. 

The Pipeline

To create the modified imagery, I utilized a ComfyUI (Stable Diffusion) pipeline. By layering IP-Adapters and ControlNet (Canny and Depth), I was able to maintain the structural integrity of the original photographs while "skinning" them with synthetic textures.

The animation was driven by an AnimateDiff workflow that allowed for shifting weights between ControlNets and IP-Adapters over time. This enabled a complex morphing effect where one image would dissolve into another through a third "buffer" state. By modulating the influence of the IP-Adapter against the ControlNet, I could control the tension between the physical photograph and the neural hallucination.

Technical Obsolescence as Aesthetic

Critically, this specific workflow is now a "lost" technology. Due to the rapid evolution of ComfyUI and the versioning of IP-Adapters, this exact visual output cannot be replicated.

I view this as a digital parallel to the early era of digital photography. I recall a specific Ricoh digital black-and-white camera—a low-resolution tool that produced a unique, visceral digital grain that was lost in subsequent "upgraded" models. 

This study serves as a similar timestamp: a specific, unrepeatable interaction between artist and algorithm before the software "improved" away the artifacts.