Verified against Midjourney · 2026-08-14
Keep the same model reading consistent across a fashion lookbook's different looks
A fashion-editorial workflow using --cref with a deliberately low --cw so the model's face stays recognizable across a lookbook while each look's garment and drape render on their own terms, plus explicit fabric-movement language to counter Midjourney's tendency to render clothing as static rather than worn.
The prompt
Ready to copy — highlighted parts are example details you can swap.
MODEL REFERENCE One clear reference image establishing the model's face and general build. Its URL goes in https://cdn.midjourney.com/mno345-lookbook-model.png. CHARACTER WEIGHT — KEPT LOW ON PURPOSE --cw 20 — set this low, in the 10-40 range, not the 100 default. At --cw 100, --cref pulls the reference image's clothing along with the face, which actively fights a lookbook where every look is a different garment by design; a low --cw matches face and general build while leaving the garment entirely to this look's own text description below. THIS LOOK'S GARMENT an oversized ivory raw-silk trench coat, cinched loosely at the waist, worn open over nothing beneath FABRIC BEHAVIOR raw silk with visible slubbed texture, catching a light breeze so the coat's open front lifts slightly away from the body mid-stride — name how the specific fabric should move or hang, since Midjourney's default rendering of clothing tends toward static, catalog-flat drape unless the movement is stated explicitly; a silk slip dress and a structured wool coat should visibly behave differently in the same gust of wind or the same stride, and the model will not supply that difference on its own from the garment name alone. POSE AND ENERGY mid-stride on a concrete runway, caught in motion rather than posed still EDITORIAL MOOD austere, high-contrast studio editorial, minimal set, all attention on the garment WHAT NOT TO DO Do not describe the garment as "flowing" or "elegant" as a substitute for naming the actual fabric and its behavior — those words describe an intended feeling, not a physical property the model can render, and tend to default back to the same generic drape regardless of which garment is named. PARAMETERS --cref https://cdn.midjourney.com/mno345-lookbook-model.png --cw 20 --ar 3:4 --v 7 OUTPUT an oversized ivory raw-silk trench coat, cinched loosely at the waist, worn open over nothing beneath, raw silk with visible slubbed texture, catching a light breeze so the coat's open front lifts slightly away from the body mid-stride, mid-stride on a concrete runway, caught in motion rather than posed still, austere, high-contrast studio editorial, minimal set, all attention on the garment --cref https://cdn.midjourney.com/mno345-lookbook-model.png --cw 20 --ar 3:4 --v 7 ACROSS THE FULL LOOKBOOK Reuse the identical https://cdn.midjourney.com/mno345-lookbook-model.png for every look in the set, and re-check --cw each time a new look's garment silhouette is very different from the reference image's original outfit — a look far more fitted or far looser than the reference tends to need --cw pushed toward the lower end of the range to keep the reference from fighting the new silhouette.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
A fashion lookbook creates a specific conflict that a comic-panel or portrait use of --cref does not: every look is, by definition, a different garment, so the reference image's own clothing is actively unwanted information in every generation after the first. --cw 100, the default, matches face, hair, and clothing together because Midjourney has no separate lever for "match the face but not the outfit" built into the parameter itself — the only way to get that selective match is to deliberately lower --cw into the 10-40 range, which biases the reference match toward facial features and general build while letting the new look's own text description take over the garment entirely, rather than having the reference outfit bleed through or visibly compete with the described one.
Naming actual fabric behavior instead of feeling-words like "flowing" or "elegant" targets a specific, well-known rendering bias: Midjourney's default clothing rendering tends toward a flat, catalog-still drape because that is how a large share of its fashion-photography training data actually presents garments — folded, hung, or standing still rather than caught mid-movement — so a word like "elegant" gives the model an emotional target with no physical information about how the material actually behaves, and it falls back to that same static default regardless of which garment is named. Describing how a specific fabric moves — raw silk lifting in a breeze versus a structured wool coat holding its shape through the same stride — gives the model an actual physical difference to render between two different garments, which "flowing" and "elegant" alone cannot supply since neither names a material property at all.
Re-checking --cw whenever a new look's silhouette differs sharply from the reference image's original outfit addresses the same style-thinning dynamic seen in other reference-image workflows: the bigger the visual gap between what the reference image actually shows and what the current generation is describing, the harder the reference tries to reassert its own visual information, so an oversized, loose silhouette following a reference photographed in a fitted outfit typically needs --cw pushed even lower than the rest of the lookbook to keep the reference's original clothing from fighting the new look's shape.
What you get back
A mid-stride editorial shot of a model whose face and build clearly match the reference image, wearing an oversized raw-silk trench with visible texture and a front hem lifting naturally in motion, against a stark high-contrast studio backdrop — with the same face carrying recognizably into other looks in the set that swap the garment entirely.
Verified against
Midjourney v7 · 2026-08-14
Changelog
- 2026-08-14 — Initial publish, verified against Midjourney v7 --cref/--cw for consistent-model fashion lookbook sequences.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
