Lab / [03] — IndSurrealStyle LoRA[2026]
teaching a
machine my
darkroom
IndSurrealStyle started in the darkroom: composite prints of blindfolded figures, black ovals on taut wire, and power lines running through industrial ground. I wanted to know whether a model could learn that visual language without flattening it. I curated the dataset, wrote and restructured the captions, and ran the training on Civitai, testing base models until the outputs started to feel like mine. This page documents the pipeline, including the parts that failed.
01source
[ replace — source print ]
01020304
component · 01
dots & spots
black ovals strung on taut wire
component · 02
lines
power lines, wire, tension across the frame
component · 03
distorted figure
veiled, blindfolded — body as conduit
component · 04
industrial ground
pylons, fields, infrastructure
fig. 01 — one print, broken into the four elements that recur across the whole body of work.