The direction language

Pick a named posture and the model stands the way you asked. Every posture in the library carries a real photograph and a body of physical phrases — where the weight sits, what the hands do, which way the head turns — because adjectives do not survive contact with an image model.

Ask for "relaxed" and you get a mannequin. Not because the model misunderstood, but because it understood perfectly and had nothing to go on: an adjective leaves every joint unstated, and unstated joints get filled with defaults. Defaults are what a mannequin is.

So the lookbook stopped taking adjectives. You pick a posture — a named one, with a photograph on the card — and behind that card sits the thing the model actually needs: physical phrases about weight, hands, feet, torso and head.

The direction language →

What a posture is

Four things, and none of them is a mood:

Where the weight sits. Which leg carries it, or whether she is between legs entirely.
What the hands are doing. In a pocket, at a hem, holding a strap, hanging.
Whether the feet are planted or mid-stride. These photograph completely differently.
How far the torso and head are turned off the lens. Separately — they are not one number.

Every one of those is checkable by looking at the result, which is the whole point. "Is the weight on the back leg" has an answer. "Is she relaxed" does not.

Why the card shows a photograph and not a diagram

Because we measured it, over the same identity and the same four postures, and it was not close: the photograph scored 10 of 12 geometry points; the pose diagram scored 4.5.

The diagrams did not fail at the drawing step, they failed at the derivation step. One came back mirrored. One came back as a generic front-facing pictogram with no turn in it at all. And the model then copied those errors faithfully — which is exactly what a good model should do with a bad instruction.

The frame you see on the card is the frame the model gets. There is no second interpretation between you and the result.

Naming your own

The library ships with the postures a lookbook actually uses — Contrapposto, Mid-Stride, Over-the-Shoulder, The Lean. You are not limited to them. A posture you save becomes yours: it carries your reference photograph and your phrases, and it appears beside the house's own the next time you build a look.

This matters more than it sounds. A house has a way its women stand. That way is not in any general library, and it is the difference between frames that look like your brand and frames that look like everyone's.

What the judge checks, and what it does not

Every posed frame is read before you see it, and the read is graded against the posture's own phrases — the weight, the hands, the feet, the turns. When one is missed, the miss is named for the retry, in the language of the body rather than of the software.

It does not grade carriage, and that is deliberate rather than unfinished. A rubric that asked about the spine, the stack of the head and the line of the chin shipped and was withdrawn: we tested every model in the production lane against posed frames, and not one read carriage reliably — the best managed it in none of four attempts.

A question no reader can answer does not return "I don't know". It returns a confident guess, and a confident guess here would spend your tokens redrawing a slump that was never in the frame. Five things graded truthfully beat six graded hopefully.

What it costs

Nothing extra. Choosing a posture is part of composing a look — it changes what the develop asks for, not how much of it there is.