Can AI put my clothes on a model from a flat lay?

Yes. A generator reads the garment from your photo and draws it on a model from a reference photo, in the pose and scene you choose. The result is a new drawing of the garment, so check colour, print, length and hardware against the real piece before you post.

Is a flat lay or a ghost mannequin photo better?

Both work: Claid and Veeton, for example, recommend either. A ghost mannequin or hanger photo shows some of the garment's shape, which a flat lay cannot, so for pieces that sell on drape or volume it gives the generator more to go on.

Why is the hem length wrong on the AI model?

Because a flat lay does not show where the hem sits on a body, so the generator decides. In our own output a dress described as midi came out ankle length. Judge length on full-length shots against the measured garment.

Why is the model wearing shoes I did not choose?

Anything you do not supply comes from the model reference or the generator's guess. Dress the reference in the shoes and trousers you want, or supply them as images if your tool accepts several.

Can I do this in ChatGPT or Gemini?

Yes. Attach the garment and a model photo and describe the shot. It works well for a few images; for a weekly catalog the cost is re-attaching the references and re-checking the face and the garment on every image.

Do I have to say the photo is AI-generated?

It depends on where you sell and on each platform's rules. Our guide to labelling AI fashion photos in the EU covers the AI Act's transparency rules and how to decide image by image.

Guide

From flat lay to on-model photo with AI, step by step

By The AURA Studio team. Published 2 October 2026.

To turn a flat lay into an on-model photo with AI, give a generator three things: a clean photo of the whole garment, a reference photo of the model, and the framing you need. It redraws the garment on the person rather than pasting it, so check every result against the real piece for colour, print, length, volume and hardware, and fix misses by changing the input.

Made in AURA from a product image of the jumper and a model reference photo. Unretouched.

What the AI does with your flat lay

A generator does not cut your garment out and lay it over a person. It reads the photo for cut, colour, print, texture and hardware, then draws the garment again on a body, in the pose, light and framing you asked for. Every part of the result comes from one of four places, and the last one is where most surprises start.

Where each part of an AI on-model photo comes from
Part of the photo	Comes from	What that means
Colour, print, cut, closures, hardware	Your garment photo	As good as the photo: anything blurred, folded under or cropped is guessed
Face, hair, skin tone, build	The model reference	One clear full-length reference keeps the same person across a catalog
Pose, scene, light, framing	Your choices	Framing decides what you can check: a waist-up crop hides the hem
Shoes, what is under a jacket, the garment's back, how the fabric falls	The reference photo, or the generator's guess	Supply it, choose it through the reference, or accept that it is invented

A flat lay leaves one more gap: it shows the garment with no body inside it. Where the hem lands on a leg, how full a sleeve hangs and how a skirt moves all have to be inferred. Hanger and ghost mannequin photos carry a little more of that shape. How to shoot the input is covered in our guide to photographing a garment for AI; this guide starts where that one ends.

Three garments, from product image to photo

Each of these used one product image, one model reference and a short description of the garment, nothing else. None was retouched.

Product image of a blush floral wrap dress with flutter sleeves and a side tie, on white — The print, the side tie and the flutter sleeves come through. The white trainers come from the model reference, because only the dress was supplied.

A blonde woman wearing the floral wrap dress and white trainers on a path under spring blossom — The print, the side tie and the flutter sleeves come through. The white trainers come from the model reference, because only the dress was supplied.

Model reference: a woman with long dark hair in a black long-sleeved top, black trousers and black flats against a grey backdrop — Model reference, product image, result. The asymmetric zip, the silver hardware and the cropped length hold. The black top, trousers and flats are the reference's own clothes.

Product image of a tan leather biker jacket with an asymmetric zip and silver hardware, on white — Model reference, product image, result. The asymmetric zip, the silver hardware and the cropped length hold. The black top, trousers and flats are the reference's own clothes.

Product image of a cream chunky cable-knit roll-neck jumper, on white — The cable layout, the roll neck and the deep ribbing hold at this size.

A dark-haired woman in the cream cable-knit jumper leaning on a cafe counter by a window — The cable layout, the roll neck and the deep ribbing hold at this size.

How these were made: the product images are catalog-style shots on white that we generated for our demo set, styled like ghost mannequin photos. They are not phone flat lays of real stock, and inputs this clean are the best case. A phone flat lay leaves the generator more to infer, which makes the checks below matter more, not less.

Step by step

Photograph the whole garment
A flat lay, hanger shot or ghost mannequin photo, straight on, in even daylight, with nothing folded under. Our garment photography guide has the method. If your tool takes extra images, add a close-up of any print or hardware.
Decide what else the model wears
Your photo covers one piece. Everything else comes from the model reference or the generator's guess. Either upload the other pieces, where the tool takes several images, or dress the reference in the trousers and shoes you want to see.
Choose the model once
Use the same reference for every garment so the catalog shows one person. Our guide to keeping one model across photos explains how, and why a real person's photo needs their written consent.
Frame for the garment
Full length for dresses, coats and trousers, so the hem is on view. Waist up for tops. A close crop for texture. A crop that hides the hem also hides a miss you need to see.
Make several, keep the best
Vary the pose or the scene and generate a handful. Most tools charge for every attempt, so count the photos you would actually post, not the attempts.
Check each one against the real piece
Side by side, at full size, with the checklist below. Compare with the garment itself, not only with the flat lay, because the flat lay may already be off in colour.
Fix the input, then regenerate
Retrying the same inputs gives you a different result, not necessarily a more faithful one. A drifting print needs a sharper photo or a close-up; a wrong colour needs daylight; wrong shoes need a different reference.
Size it and label it
Export at the size of the placement; our Instagram and TikTok sizes guide has the numbers. If you sell in the EU, read our guide to labelling AI fashion photos before you publish.

What to check before you post

Checks for an AI on-model photo, made against the real garment
Check	How it goes wrong	Fix
Colour	Shifts with the scene's light: a sunset warms the fabric too	Compare with the garment in daylight; use neutral light for the product page shot
Print and pattern	Motifs change scale or repeat, or drift into a similar print	A sharper, straight-on input and a close-up of the pattern
Length	The hem lands longer or shorter on the body than on the real piece	A full-length shot, compared with the measured length
Volume	Full sleeves, skirts or collars come out flatter or fuller	A hanger or worn photo, and a pose that shows the feature
Closures and hardware	Buttons, zips and plackets move or change in number	Count them against the garment
Logos and printed text	Letters change shape or spelling	Read them letter by letter, or add text as a separate layer after generation
What you did not supply	Shoes, undershirts or trousers that do not suit the piece	Supply them, or change the reference
The person	The face drifts from the reference, or the hands look wrong	Regenerate; a clearer reference photo helps the face

Four misses from our own output, and what fixes them

These came out of the same pipeline as the examples above, from the same kind of inputs. None was retouched. Each one is a miss you would catch with the checklist, and each traces back to something the inputs did not show.

The hem came out longer. The emerald slip dress is a midi, and its description said so in words: "midi length" was one of the tags sent with it. Both results put the hem at the ankle or below. The product image never shows where the hem sits on a leg, so the generator decided, and the written tag did not overrule it.

Product image of an emerald satin slip dress with a cowl neck and a midi-length hem, on white — A midi slip dress, then two results with the hem at the ankle and below. The black flats in the middle shot come from the model reference.

A dark-haired woman in the emerald slip dress in a hotel lobby, the hem at her ankles, wearing black flats — A midi slip dress, then two results with the hem at the ankle and below. The black flats in the middle shot come from the model reference.

The fix: judge length on a full-length shot against the measured garment, and put the length in the product copy. If no result is close, do not use that shot to sell the fit.

The undershirt changed. The navy suit was supplied as a jacket and trousers with nothing underneath. In the boutique shot the model wears the grey T-shirt from his reference photo. In the night shot, from the same inputs with a different scene, he wears a dark one.

A dark-haired man in a navy double-breasted suit over a grey T-shirt, with white trainers, in a menswear boutique — Same suit, same reference, two scenes. The T-shirt under the jacket was never supplied, so it changed between generations.

The same man in the navy suit jacket over a dark T-shirt on a neon-lit street at night — Same suit, same reference, two scenes. The T-shirt under the jacket was never supplied, so it changed between generations.

The fix: anything that must stay the same across a set has to be in an input. Shoot the reference in the shirt you want, or supply the shirt in a tool that takes several garment images.

The sleeves lost volume. The terracotta blouse has full sleeves gathered into deep cuffs. On the model they read slimmer, partly because the crossed-arms pose presses them against the body.

Product image of a terracotta silk blouse with a collar, a concealed placket and full gathered sleeves, on white — Full sleeves on the product image, slimmer on the model, with crossed arms hiding most of the volume. Colour, collar and placket hold.

A woman with short black hair in the terracotta blouse, arms crossed, on a rooftop at golden hour — Full sleeves on the product image, slimmer on the model, with crossed arms hiding most of the volume. Colour, collar and placket hold.

The fix: choose a pose that shows the feature that sells the piece, arms relaxed for sleeves and walking for a skirt. A hanger or worn photo also tells the generator more about volume than a flat surface can.

The shoes came from the reference. A charcoal wool overcoat on a snowy road, worn with white trainers. Nobody chose the trainers: they are in the model's reference photo, and the coat was the only garment supplied. It is the same lesson as the T-shirt, and our consistent-model guide shows it with a suit.

In a general AI model or a fashion tool

General image models can do this. Attach the garment photo and a model photo, ask for the garment on that person, and say what else they should wear. Google's documentation says its image models can mix up to 14 reference images in one request, depending on the model, so a whole outfit can go in at once (Gemini API docs). The cost is repetition: you re-attach the references, restate the instruction and re-check the result for every image. Our comparison with ChatGPT and Gemini covers where that stops paying off.

Fashion tools store the model and fix the output sizes for you, and they differ in what they take as input. Here is what five of them say on their own pages, next to ours:

Claid: Flat lays and ghost mannequins "work best" and hanger shots "work well"; complementary items can be added to style a full outfit
Nightjar: A flat lay, hanger shot or ghost mannequin photo, or a product link; close-ups of prints, trims and hardware can be added
Photoroom: One clothing photo, or several clothing and accessory images to build a complete outfit
Veeton: Recommends ghost mannequin or flat lay photos; phone photos work if the garment is fully visible
WearView: Any clothing photo: flat lay, hanger, or on a person
AURA Studio (us): One garment image per photo, as PNG, JPEG, WebP or AVIF up to 12 MB; everything else comes from the model reference

Checked 2 October 2026 on Claid, Nightjar, Photoroom help, Veeton, WearView, AURA.

AURA's one-garment limit is the catch to know about: a top and trousers only go together if they are in the same picture, so the reference photo does more work. For prices, video and the catch with each tool, see the best AI fashion photography tools in 2026. To run this workflow on your own garments, start a free week: it covers 10 photos and one video, with no card.

When a flat lay is not enough

Sheer, sequinned and high-shine fabrics, which change with the light on them and behind them. Our garment guide covers what to add for the hard cases.
Fit claims. An on-model photo shows a plausible fit on one body, not a measured one. Size charts and measurements stay the source of truth.
Campaign images. When one picture has to carry a season, a shoot still makes the better image; our breakdown of what a shoot costs shows where the two meet.

Sources

Questions

Can AI put my clothes on a model from a flat lay?: Yes. A generator reads the garment from your photo and draws it on a model from a reference photo, in the pose and scene you choose. The result is a new drawing of the garment, so check colour, print, length and hardware against the real piece before you post.
Is a flat lay or a ghost mannequin photo better?: Both work: Claid and Veeton, for example, recommend either. A ghost mannequin or hanger photo shows some of the garment's shape, which a flat lay cannot, so for pieces that sell on drape or volume it gives the generator more to go on.
Why is the hem length wrong on the AI model?: Because a flat lay does not show where the hem sits on a body, so the generator decides. In our own output a dress described as midi came out ankle length. Judge length on full-length shots against the measured garment.
Why is the model wearing shoes I did not choose?: Anything you do not supply comes from the model reference or the generator's guess. Dress the reference in the shoes and trousers you want, or supply them as images if your tool accepts several.
Can I do this in ChatGPT or Gemini?: Yes. Attach the garment and a model photo and describe the shot. It works well for a few images; for a weekly catalog the cost is re-attaching the references and re-checking the face and the garment on every image.
Do I have to say the photo is AI-generated?: It depends on where you sell and on each platform's rules. Our guide to labelling AI fashion photos in the EU covers the AI Act's transparency rules and how to decide image by image.