Before Words: Dry Roof Loft — AI Portrait Prompt

Use the attached reference photo as the identity reference: keep the same face, skin tone, hair and features as the person in that photo, without beautifying, aging or altering them. Create a square 1:1 museum-catalog frame with a strong first read and a slower second read at exactly 2880x2880 pixels. Location: a dry roof loft where cool air enters one louvered vent. Show the subject in a faded slate work shirt buttoned to the second button, opening a louver a finger's width with one hand. Make one rigorous physical condition the conceptual engine: a clock-driven barograph drum whose pen draws one unbroken line as a sealed bellows breathes with the passing air turns a private condition into something two distant rooms can read together. Explore what a slow drift becomes once someone builds a way to watch it; direct expression and posture toward lucid attention, never melodrama or ad-like confidence. Use an eye-level 50mm viewpoint with natural perspective and one plane of critical focus. Make the composition leave honest vertical distance between the sensing point and their hands. Render it as conceptual house-systems photography built from functioning telegraph hardware, accurate liquid-column behavior, restrained interior direction, and editorial portrait fidelity, using flat overcast illumination that keeps every glass column readable without glare and a palette of lead grey, unpolished tin, and one stripe of verdigris. Photorealistic: this must read as a real photograph from a physical camera, never an illustration, painting, or 3D render. Resolve real surface detail: fine grit in the texture of concrete, efflorescence at damp edges, and hairline cracks that follow stress. Keep one identity-locked person, credible hands and fingers, natural fabric, and coherent scale, contact, reflections, shadows, airflow, deformation, and materials. Background people only when essential, always secondary and clearly different. Avoid steampunk goggles, glowing holograms, dripping food props, anatomical organ diagrams, and legible dial lettering anywhere. No text, labels, logos, signatures, watermarks, random symbols, plastic skin, excessive bloom, or glossy AI finish.