Series / AI-assisted portrait / 2026
Head Reconstruction
Portraits rebuilt as systems of contours, folds, masks, grids, and unstable generated surfaces.
Series / Surface as structure
Head Reconstruction treats the portrait as an editable structure rather than a fixed likeness. Facial surfaces are measured, divided, folded, masked, and rebuilt until anatomy begins to read like an uncertain map between skin, sculpture, drawing, and machine vision.
The series begins with precise interventions in Photoshop: line systems, graphic masks, contour divisions, and fragments are composed directly over the head. These marks become structural instructions rather than decoration.
The image then moves through ComfyUI, where the original portrait and its drawn construction are regenerated together. Repeated passes preserve the pose while allowing material, identity, and facial architecture to shift.
Process / Photoshop to ComfyUI
The workflow is circular rather than linear. Photoshop establishes the drawing, masks, proportions, and image hierarchy; ComfyUI interprets those decisions as material and form. Selected generations return to Photoshop for correction and recomposition before entering another controlled generation pass.
Read the Photoshop to ComfyUI workflow →