AI fashion models for your store: where they work, and where they don't

Every new drop starts the same way. To sell it you need photos, and decent photos need a model, a location and someone behind the camera — for a small shop, that bill comes back every season. So when you hear that software can produce a model from a sentence, the first instinct is to try it right away. The better first question is how far you can take it before it causes trouble.

What a text-to-image model can and can't show

The Vipix tool works like this: you describe who's in the picture, what they're wearing, where they're standing, the light and the style, and it builds an image of a person in that setting that looks like a photograph. Generate photorealistic images is tuned for exactly that — people, portraits, ad models and brand images. If what you need is an object on its own or a scene with nobody in it, Generate images is the better fit.

One fact shapes every limit that follows: it generates from text, and it doesn't take your photos. There's no upload box. It never sees the jacket in your stockroom, and it never sees anyone's face. It reads your description and builds an image to match, from nothing.

That puts it in a different category from most tools you'll find when you search for AI fashion models, which ask you to upload a photo of your garment and dress a generated model in it. With a text-to-image tool, the clothes in the picture are invented to fit your words — seams, fabric, zipper, label, the way a hem falls. Even the upload-your-garment tools have to rebuild drape and fit on a body that was never photographed wearing the item. Either way, the result isn't a photo of what your customer will receive.

The clearer the description, the closer the result: who, the setting, the light and the style beat one vague line. Each run produces a new image rather than repeating the last one, so pressing Generate again is the normal way to get variations. AI results can contain mistakes, so check every image before it goes into a campaign — our guide on what to check in an AI portrait before you post it covers what to look for.

Where AI models belong in a store

The setup with the least risk has two layers. Real photos hold the product page, where customers decide. AI images go around it.

Banners, store covers and mood shots

A store cover image, a banner for a sale, an illustration for a post, the opening image of a seasonal collection: places where the image sets a mood rather than stands in for the item. That's the job a generated model does well.

A brief for your photographer

This use gets overlooked. Mock up the scene you have in mind — the setting, the light, how the model stands — and show it to your photographer before the real shoot. A mock-up talks to someone who knows it's a sketch and will reshoot it with a real camera; a product photo talks to a customer who has only that photo to trust. Same image, two very different readers.

Not on the product page

The product page is where someone buys on the strength of the photo, so it should show what they'll actually get. Returns are already common in online retail: in its 2025 returns report, the National Retail Federation said retailers estimated that 15.8% of their annual sales would be returned, and that "an estimated 19.3% of online sales will be returned in 2025." The report doesn't blame photos for that, but a photo that doesn't match the item only adds to the risk.

One brand that planned to test AI models framed them the same way. When Levi Strauss & Co. announced a partnership with Lalaland.ai on March 22, 2023, it said it was planning tests "using AI-generated models to supplement human models." An editor's note added on March 28 went further: "we are not scaling back our plans for live photo shoots, the use of live models, or our commitment to working with diverse models." Supplement, not replace.

New US rules for AI people in ads and listings

This isn't legal advice, and rules differ by state and by platform. Two are worth knowing if you sell in the US.

New York's synthetic performer law

New York now requires ads to say when they use a "synthetic performer." The bill (A.8887-B / S.8420-A) defines one as a digitally created asset made "using generative artificial intelligence or a software algorithm" to create the impression that it is performing as a human performer "who is not recognizable as any identifiable natural performer." Anyone in business who produces or creates an ad for their goods or services "shall conspicuously disclose in such advertisement that a synthetic performer is in such advertisement, where such person has actual knowledge." Violations carry civil penalties. There are carve-outs, including ads for expressive works such as films and video games when the performer is used the same way as in the work, audio ads and AI used only to translate a real performer's language.

Governor Kathy Hochul signed it on December 11, 2025, and it took effect 180 days later, on June 9, 2026 — the date set out in a January 2026 client alert from the law firm Cooley and confirmed in an August 2026 alert from Croke Fairchild Duarte & Beres, which says the law "took effect on June 9, 2026." For a store owner the practical reading is short: if an ad for your products shows a generated person who looks like a real human performer, and you know it, the law expects you to disclose that conspicuously in the ad itself.

Amazon's tag for AI-generated people

In a seller forum announcement posted in late July 2026, Amazon told sellers: "You must now tag product images that contain photorealistic AI-generated people with specific metadata before you add them to your listings and A+ content in any of our worldwide stores." The tag is a keyword, "contains-synthetic-performer," added with an IPTC-compatible metadata editor to the dc:subject (XMP) field of the file; Amazon says it will then show customers an indicator that the image includes AI-generated people, where applicable. Images that "Only features real people (even if they have been altered with AI tools)" don't need it. Amazon gives the reason in the same post: "Recent legislation requires disclosure when images or videos in advertisements contain photorealistic AI-generated people."

One practical detail: Vipix doesn't add that keyword to the files you download, so if you list on Amazon, tagging is your step. Whether AI images belong in your listings at all is a separate question; check your marketplace's own image rules before you upload.

A person who doesn't exist vs. a real person's face

Works: generating a person who doesn't exist to wear your kind of clothing in the setting you need. That's what the tool is for — a face that belongs to no one, made for that one image, standing in for no real person.

Doesn't: aiming at the face of a real person, whether that's someone you know, a model, an influencer or a celebrity.

What makes the difference is whether someone's identity is being used. A person who doesn't exist has no identity to exploit: nobody gets tied to something they didn't do, and nobody has to explain a photo they never posed for. When an image targets a real person, what's being used isn't data anymore; it's that person.

New York's rules for fashion work draw the same line from the other side. Under the Fashion Workers Act, in effect since June 19, 2025, the state Department of Labor explains that a model's digital replica — "a computer-generated or artificial intelligence-enhanced representation of a model" — requires the model's "separate and explicit written consent," and that "The model must approve in writing the scope, purpose, rate of pay, and length of time the replica will be used."

Two product facts back this line up. There's no way to upload a face: the tool works from text only. And the tool's own FAQ says it's meant for images of adults in commercial settings, and that the face of a specific real person or a celebrity won't come out as an exact likeness — it's built to create people for commercial scenes, not to copy a face. Don't use it to aim at someone real.

What clothing you can show

Can show: everyday clothes, workwear, evening wear, activewear — and lingerie, sleepwear, swimwear and bikinis. Those are big categories, and sellers in them need photos like everyone else. They're within the normal scope of the tool, not an exception you have to ask for.

The yardstick is the community standard of the platform where you sell. For a US store, the practical question is: would this image be fine as a regular post on Instagram or Facebook, and does it fit the photo rules your marketplace publishes? Amazon, for instance, publishes its own rules for apparel images; our roundup of Amazon's apparel photo rules walks through them. If the answer is yes, the image is within what the tool is for.

Can't show: nudity and sexual content. That line has no exceptions.

What makes the difference is the platform's community standard — not the taste of whoever wrote this article, and not a separate list Vipix made up. Setting the bar there helps you directly: the line sits where your posts have to stand anyway, so you don't plan a whole set around images you couldn't post where you sell.

When a description gets refused

One more distinction, because mixing it up costs you. A description that gets refused is not the same as content above the line. A safety filter can occasionally catch a harmless description because of how it happens to be worded; rewording that same harmless description is fine. Content above the line is different: no wording opens it, because the problem isn't the words.

So if a description of an adult model in ordinary swimwear is refused once, don't conclude that your whole category is off-limits and drop it. Describe the same harmless shot in plain words and try again.

The mark in the corner

Every image from Generate photorealistic images carries a small Vipix mark — the text vipix.vn — in the bottom-right corner, and today it stays there on every plan, paid ones included; there's no setting to turn it off.

The reason is simple, and we think it's fair: a picture of a person who looks real but was never photographed should say what it is. Without any mark, it's easy to pass off as a real photo. The corner mark ties the image to the tool that made it, and it makes the image harder to pass off as a real photo or as "proof" of something that never happened.

In practice it splits along the same line as the rest of this article. On banners, covers and mood shots, a small corner mark gets in nobody's way. On product detail images, it would be noticeable — but those were never the place for a generated image anyway.

Don't count on the mark as your disclosure, though. It's there for the reason above; it wasn't designed around New York's ad rule, and it isn't the metadata tag Amazon asks for. Those are steps you take as the advertiser and seller.

Frequently asked questions

Can I put my own clothes on an AI model with this tool?

No. Generate photorealistic images works from a text description only. There's no way to upload a photo of your garment, so the clothes in the image are invented to match your words.

Do I have to disclose AI models in my ads?

In New York, yes: since June 9, 2026, an ad that includes a synthetic performer must disclose it conspicuously when the advertiser knows. Amazon also requires a metadata tag on product images that contain photorealistic AI-generated people. Rules elsewhere can differ, and this isn't legal advice.

Can I make an AI version of a real model or influencer?

Don't use this tool to aim at a real person. In New York, the Fashion Workers Act requires a model's separate, explicit written consent before their digital replica is created or used.

Can I show swimwear or lingerie on an AI model?

Yes. Swimwear, lingerie, sleepwear and bikinis are ordinary product categories here, judged by the community standards of the platform where you'll post. Nudity and sexual content are not.

Does the Vipix mark come off on a paid plan?

No. Today every image from Generate photorealistic images carries the small mark in the corner, whatever plan you're on.

Try it on your next banner

Pick one image you need for your next drop — a store cover, a sale banner, a mood shot — then describe it and see what Generate photorealistic images makes of it. The cost is shown in the editor before you run it. Keep your product pages for real photos.