ComfyUI

Node-based Stable Diffusion workflows in ComfyUI, used to train a custom fashion LoRA and generate concept stills across cities and styles. Each look starts with a base prompt and gets refined through ControlNet passes, IP-Adapter style transfer, and upscaling chains.

The series below explores fashion editorials across Tokyo, Detroit, Mexico City, and Berlin, playing with how AI can sit between street-style photography and stylized concept art.