Can AI grade my garment to other sizes from one photo?

Not in any image model we know of. It resizes a drawing; it does not apply your grade rules. Use your pattern measurements for each size, and treat the image as an illustration of them, checked against a real sample.

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

Where a platform or law requires it, yes: Amazon’s contains-synthetic-performer tag, and New York’s disclosure rule for synthetic performers in ads. Where it is not required, a short line next to the size information costs little and answers the question before a shopper asks.

Is one model plus a size chart enough?

It can be, if the chart is complete and the model’s height, measurements and size worn sit under the photos, since Baymard’s participants used that combination to choose a size. Extra sizes help most for shoppers whose bodies differ most from the model, so add them where your returns show fit problems.

How to show clothes on different body sizes with AI

Checked September 24, 20265 min read

Short answer

Only show a garment on a body size you have real fit information for. Measure the garment in each size you will show, give the model those measurements and the size, label every photo with the size worn and the model’s height, and keep at least one real fit photo per size group. AI can make a plus-size or petite model look convincing while inventing how the garment fits, and a misleading image can break US truth-in-advertising law.

Why shoppers want to see their size

Baymard Institute’s 2026 survey of 1,922 US online shoppers found that 24 percent of women and 14 percent of men who buy apparel call themselves plus-size, and 16 percent of women call themselves petite. Size accuracy and fit details were the information those shoppers looked for most in reviews, at 48 percent. In Baymard’s usability tests, shoppers compared the model’s height, measurements and size worn with their own before choosing a size, and 83 percent of desktop apparel sites did not give enough sizing information.

Brands have tried to close this gap for years. Levi’s wrote in 2023 that photo shoots are generally limited to one or two models per product. In January 2020 ASOS tested See My Fit with the AR company Zeekit, digitally mapping 800 dresses onto 16 models in sizes 4 to 18.

Where AI goes wrong on size

Grading, the pattern maker’s job of scaling a garment from one size to the next, is not something an image model can do. Ask for your size 8 dress on a size 18 model and it draws a size 18 body in a dress that looks like it fits, because it has no information about how your dress fits at that size. It cannot know that your size 18 keeps the same bust dart, runs longer at the hem, or pulls across the hips. The result can look better than the real garment, and that is the problem.

What the AI does	What the shopper concludes	What arrives
Scales the garment to the body without changing its proportions	The larger size is cut the same way	A larger size that may be graded differently
Removes pulling, gaping and riding up	It sits smoothly at my size	A garment that pulls where the photo showed none
Keeps the hem at the same point on a shorter or taller body	The length suits my height	A hem in a different place
Draws stretch where the fabric has none	It will give over my hips	A fabric that does not give

The honesty rule

In the US, the Federal Trade Commission requires ads to be truthful and not misleading. It judges an ad by everything it communicates, pictures included, and treats it as deceptive if it is likely to mislead a reasonable consumer about something that matters to the purchase. How a garment fits at a given size matters to the purchase. More in FTC rules for AI product images.

Platforms set the same bar. Amazon’s product image guide says the main image must represent the product accurately, including its real scale, quantity and color. Google says its shopping try-on shows only model sizes consistent with the sizes a brand actually sells.

An honest workflow

Choose the sizes to show, for example your smallest, a middle size and your largest, or the sizes that sell most.
Get real fit data for each: measure the garment flat in each size (chest, waist, hip, length, sleeve) and, if you can, photograph a fit model wearing it.
Cast one approved model per size group, with a height and build that match a real person of that size. See how to cast AI models responsibly.
Write the size and its measurements into the prompt, for example "wearing size [18], garment chest [124 cm], length [98 cm], relaxed through the hip, hem [just below the knee] on a [5 ft 6 in] model."
Render, then compare each photo with the fit photo or with the real garment on a person of that size. Reject any image that fits better than the real thing.
Label every image with the size worn and the model’s height, and say the model is AI-made where a platform or law requires it.
After launch, watch returns and fit reviews by size. In January 2024, for example, Amazon described a Fit Insights Tool that gives brands an analysis of returns and customer feedback about fit.

Fit words and landmarks for the prompt are in how to show fabric texture and fit.

What to put under the photo

Size and disclosure lines

Model is [5 ft 10 in (178 cm)] and wears size [M].
Model image generated with AI from photos of this garment in size [M].
Garment measurements for every size are in the size chart.

Give height, measurements and size worn, because those are the numbers Baymard’s test participants compared with their own. If the model is a digital twin of a real person, say so, for example "Digital model of [name], used with permission", and keep the signed consent on file. Amazon’s synthetic-performer tag is for people made entirely by AI, not for images based on a real person.

When to shoot real people instead

Swimwear, lingerie and shapewear, where the fit is the product
Stretch denim and body-con pieces, where the body changes the garment’s shape
Adaptive clothing, where the wearer’s body and mobility aids matter
Any size you have never fitted on a person

For the rest, a hybrid works: real fit photos for each size group as the record of how it fits, and AI photos for styling and scenes around them. Before anything goes live, run the AI image QA checklist.

Questions people also ask

Can AI grade my garment to other sizes from one photo?: Not in any image model we know of. It resizes a drawing; it does not apply your grade rules. Use your pattern measurements for each size, and treat the image as an illustration of them, checked against a real sample.
Do I have to say the model is AI-generated?: Where a platform or law requires it, yes: Amazon’s contains-synthetic-performer tag, and New York’s disclosure rule for synthetic performers in ads. Where it is not required, a short line next to the size information costs little and answers the question before a shopper asks.
Is one model plus a size chart enough?: It can be, if the chart is complete and the model’s height, measurements and size worn sit under the photos, since Baymard’s participants used that combination to choose a size. Extra sizes help most for shoppers whose bodies differ most from the model, so add them where your returns show fit problems.

Where Overs fits

Overs casts one model per character sheet, and earlier sheets can be reused as a talent library, so you can keep a separate approved model for each size group. It works from reference photos and does not train on your product, so the fit you show still has to come from your own measurements and fit photos.

Free for 40 photos a month. The AI that makes the photos is billed separately, on your own key, with no markup from Overs.

Start free How it works

Sources

Checked on September 24, 2026. Prices, specs and rules change; follow the links for the current versions.