TL;DR
If a generated model’s legs look short or the build feels off,
rewriting the prompt with “longer legs” or “better proportions” rarely fixes it.
Proportions are shaped mostly by two other things: the model’s face and the fit shown in your garment photo.
When you have a specific build in mind,
preparing the right face is the fastest route.
When the Proportions Look Off
You upload a garment shot, generate a lookbook,
and something about the body doesn’t match what you pictured.
The legs might read shorter than in the reference photo.
The overall build might feel younger or heavier than your brand’s usual look.
The instinct is to fix it in the prompt —
add “long legs,” “slim proportions,” “tall model” — and generate again.
That rarely moves the needle much.
Proportions get set upstream of the prompt, by what’s already in your inputs.
What Actually Decides Proportions
So if the prompt isn’t shaping the body, what is?
Two things carry most of the weight.
The first is the model’s face.
The model reads facial cues — age, bone structure, overall impression —
and infers a body to match.
A youthful, rounder face pulls toward a shorter, softer build;
a face styled for a mature look pulls toward adult proportions.
This is part of why rewriting just the prompt rarely helps —
the face has already told the model what kind of body to build.
The second is the garment photo itself, when it’s an on-body shot.
If your input is a photo of someone actually wearing the item,
the proportions in that photo — how long the sleeves run, where the hem sits,
how the fit falls on the body — carry into the result.
The garment photo isn’t only a reference for the item.
It doubles as a reference for proportions.
Proportions in the Output
Here’s the face effect in practice.
We kept one on-body garment shot fixed as the outfit input and generated three times, changing only the model’s face.
The outfit never changed and the prompt said nothing about bodies,
yet the impression of build and proportion reads differently in all three.
The youthful face pulled toward a slight, willowy frame;
the mature face toward a long, upright model build; the round,
friendly face toward a petite, realistic one.
Push it further and the effect gets extreme —
feed in a child’s face and you get toddler proportions.
The face leads the body.
Now the garment-photo effect.
This jacket photo has the face cropped out of frame, leaving only the on-body fit.
The sleeve length and the way the jacket sits on the body track closely with the input photo — compare where the sleeves end in each.
Nothing in the prompt specified proportions; the input photo did that work.
When You Want a Specific Build, Start With the Face
So what about the reverse — when you already have a clear build in mind,
like a tall, slim adult model?
The same logic applies: prepare a face whose impression matches the build you want,
and let it lead.
You create and keep those faces in My Assets > Models.
If you need a grown, mature look, build a face that reads that way;
if you want something more youthful, make one of those instead.
Rather than chasing a single perfect face on the first try,
generate a few with slightly different impressions, put the same garment on each,
and compare — the one that fits your brand’s proportions usually stands out fast.
For the full walkthrough on choosing and generating faces,
our model selection guide and the create-a-model guide cover that side in detail.
Summary
- Prompts have limited influence on a model’s body proportions.
- The model’s face is the biggest factor — it tells the AI what body type to build.
- If your garment input is an on-body shot,
its fit and proportions carry into the result. - Want a specific build?
Prepare a face with a matching impression in My Assets and compare a few.
Proportions aren’t the only thing your inputs decide.
Starting with why that jacket shot has the face cropped out,
our guide to cleaning up garment inputs covers the crop habits that cut down on generation errors.
And if you’re still finding your footing,
the solo lookbook walkthrough runs through the whole flow, from one photo to a finished set.