Case study — AI model creation & content production
Two models.Neither one exists.
A full content wave for a Levantine eveningwear label — built, cast and shot without a camera.
Case study — AI model creation & content production
A full content wave for a Levantine eveningwear label — built, cast and shot without a camera.
The brief
A renowned US marketing agency runs content for fashion brands. For a Levantine eveningwear and modest-fashion label, they needed a face — and then another one — that could carry every product in the catalogue.
The output required per product was fixed: three static posts and two reels. The catalogue was not fixed. It grows.
The problem
Casting, booking, styling, studio, retouch, and a usage licence with an expiry date on it — repeated per garment, indefinitely. Cost and lead time scale with the catalogue, and the licence runs out while the product is still selling.
Underneath that sat a second problem. The label sells to an audience that spans modest fashion and open fashion. Serving both properly means two models, two bookings, two shoot days — or choosing one and giving up half the market.
The casting
Both models were specified as sharing a Levantine look — the same cultural origin, deliberately distinct individuals. The faces had to read as belonging to the brand's actual customer, and the two of them had to read as coming from the same world.
One wears hijab. One doesn't. The same garment is shot on both, and the label covers its whole audience from a single production run.
Model 01
Layla — identity reference sheet
Model 02
Khlood — identity reference sheet
The process
Face shape, eyes, eyebrows, nose, lips, makeup register, expression, posture, hand placement, gaze, camera angle, background, lighting — documented per model and delivered for approval. Written to be marked up and returned.
Sign-off happened on the specification, not on finished images. The expensive part of an AI model engagement is discovering the face is wrong after production has run.
Each model carries a base character prompt into every generation, backed by six-panel reference sheets. A face that shifts between posts isn't a brand asset — it's a stock library.
Five images per model, delivered first, for look-and-feel sign-off. Cheap to correct at five frames. Expensive at fifty.
Three statics and two reels per product, images and motion generated from the same locked references, so the grid and the reels read as one campaign rather than two.
The output
Every garment ships the same package — three statics for the grid, two reels for the feed, all from the same locked identity.
One product wave — five delivered assets
Delivery
| Image resolution | 4K — 3840 × 2160 landscape, 2160 × 3840 portrait |
|---|---|
| Image formats | PNG with transparent-background option, plus JPEG on background |
| Video | MP4 / MOV, 24fps, Reels-ready 9:16 and square 1:1 |
| Turnaround | 48 hours from approved brief |
| Rights | Full commercial — unlimited use across all platforms |
The outcome
What it proves
Wusul Studio builds AI models and content systems for fashion brands and the agencies that serve them.
Start a project