Paired front + profile product views (matched set)

image prompt productimage-to-imageconsistencyecommerce

Result — Third pass on a 10-product set. Pass 1 drifted to three-quarter; pass 2 over-corrected to thin slivers; this one reads correctly.

When to reach for this

For generating a **matched set** of product views: one front view, then a

The prompt

The reference image shows a garment hanging on a hanger, photographed from the front. Show the EXACT SAME garment — identical colour, identical fabric, identical hanger, identical wall, identical lighting, identical cast-shadow direction and camera height — but now photographed from the side, as it would look hanging on a rail viewed along the row.

The garment is turned about ninety degrees. You see its side profile: one shoulder, one sleeve hanging down beside the body, and the body edge-on. The chest and any print on it are turned away from the camera and are NOT visible.

Important proportions: the garment should read as a substantial side view, not a thin sliver. Its width should be roughly three-quarters of the width it has in the reference image. Keep the sleeve clearly separated from the body with visible space between them, and let the hem show its side depth.

The garment hangs from the same hanger, hooked at the same point, at the same height, filling the same share of the frame. Background stays the same completely plain, flat, uniform light warm-grey studio wall #DFDCD5 with no texture and no pattern. Photorealistic editorial product photography, no text, no logo, no watermark, no people, no props.

Notes

**Two failure modes, both seen here.** Asking for "a side view" gives you a

kind: image
type: prompt
model: gpt-image-2
effort: medium
tool: image_generate
author: designer
status: proven
project: rail-store
created: 2026-10-03

first used at /home/ubuntu/rail-store/scripts/regen_profiles_mid.py