Most small fashion brands already have a flat lay of every product. It is the cheapest photo to take: lay the garment on a table, light it evenly, shoot from above. What they often lack is the on-model photo that shoppers actually want to see before they buy. This guide walks through turning a flat lay into an on-model photo with AI, step by step, using NotShot. It covers how to shoot the flat lay so the render comes out right, how to choose the model, pose and scene, what gets checked before you see the result, and what it costs per image.
A flat lay tells a shopper what the garment is: the color, the print, the shape of the collar. It does not tell them how it fits. Length, drape, where a waistband sits and how a sleeve falls only show up on a body. That is why most fashion product pages lead with an on-model image and use the flat lay further down, and why brands without a model shoot often see more questions and returns about fit.
The traditional way to get that image is a shoot: a photographer, a model booking, a studio, styling and a retouching pass, repeated for every drop. For a small catalog that cost adds up fast, see what catalog photography actually costs a DTC brand. The AI route skips the shoot. You keep the flat lay you already have, and the garment is rendered on a model from it.
If you currently shoot ghost mannequin images for the grid, the same idea applies, covered in more depth in the ghost mannequin alternative guide. A ghost mannequin photo works as the source image too.
The render can only be as faithful as the photo you give it. The garment in your flat lay is the reference that the render and the quality check are both measured against, so a clean source photo pays off directly. You do not need a studio, but a few habits make a real difference:
An existing packshot or ghost mannequin image works as well as a flat lay, as long as it shows the whole garment clearly. No professional photo is required to start.
The next choice is who wears the garment. There are two paths in NotShot.
Upload your own model. If your brand already works with a model, upload a clear, well-lit photo of them. The render keeps that person's features and proportions instead of swapping in a stock face, and the identity stays the same across every render and every drop. This is the part most brands care about once they move past a single test image: a catalog where every product is on the same recognizable person reads as one brand. More on uploading your own model and how identity preservation works. If your model comes from an agency as a comp card, the setcard import (in Beta) splits the sheet into individual shots for you.
Create a brand-owned AI model. If you have no model, you can generate one: a portrait first, then a 4K model built from it, which you reuse across every shoot. Creating a portrait costs 1 credit and building the model costs 3. You do this once, not per product.
Either way, pick one model for a whole collection where you can. Consistency across the grid is what makes AI on-model photos look like a catalog rather than a set of unrelated images.
With the flat lay and the model in place, you set up the shot.
Category matters here too. A dress, a pair of jeans and a swimsuit each have their own things to get right, such as the hem on a dress or the rise on jeans. The category pages explain what to watch for: dresses, pants, shirts, jackets and swimwear. For garment-specific guides see AI dress product photography on model and AI denim model photography.
Start the render. The finished 4K image comes back in under 10 minutes. Before it reaches you, every render goes through an automated quality check, not a sampled spot check.
The check compares the render with the garment you uploaded. The questions it asks are the ones that hurt a catalog when they go wrong: does the garment on the model match your flat lay, down to the print, neckline and hem; did the pose come out as requested; has the model's identity drifted; is there an artifact that would be obvious at full zoom. If a render falls under the bar, your credits come back automatically. You do not file a ticket for it. Full details are in how the AI judge and auto-refund work.
This is the main practical difference from generating images by hand with a general AI tool. Flat lay to model photo conversion fails in quiet ways: a print that is almost right, a pocket that moved, a logo that is slightly off. Those are exactly the errors a shopper notices after the parcel arrives. A check against the source flat lay catches them before the image goes on your product page.
The check is automated and it is not infallible. When you review a render, zoom in on the fabric, the details and the hands at full resolution before you publish. Every render is delivered at 4K, which is what keeps fabric texture and stitching sharp when a shopper zooms in (see why the 4K upscaling step matters).
If a render is close but not quite what you want, you do not have to render it again from scratch. Three optional passes work on a render you already have:
Most catalog images need none of them. They are there for the hero image of a drop, or for a product where the first render is nearly right.
NotShot charges per action in credits. A shipped 4K render is 4 credits. At the pay-as-you-go rate of about $0.50 per credit that is about $2.00 per image. On a subscription the price per render drops: the Bundle plan is $29 a month for 100 credits, about $1.16 per render, and the Studio plan at $299 for 1,500 credits brings it to about $0.80.
For a 20-product drop with one on-model image per product, that is 80 credits, roughly $40 at the pay-as-you-go rate, plus any refine passes you choose. Because a render that fails the quality check is refunded, you pay for images that pass, not for attempts. New accounts start with 50 free credits and no card, which is enough for about 12 renders to test on your own flat lays before you spend anything. The full breakdown is in what a NotShot credit costs.
NotShot is not the cheapest option, and it is worth being clear about that. Many self-serve tools sit in a band of roughly $0.15 to $0.60 per image, and some go well below that. NotShot's $0.80 to $2.00 per 4K image is at the premium end, close to tools like Botika, which justifies a similar price with human retouching.
What you get for the difference is the workflow described above: a brand-owned model whose identity stays locked across the catalog, up to 8 garments in one 4K render, an editorial mode for campaign images, and an automated check on every render with a refund when it fails. Do-it-yourself try-on tools, such as the FASHN app, are cheaper per image but leave the art direction and the quality control to you. Tools like Photoroom and WearView also offer ghost mannequin images, if that is the format you want to keep. For a wider comparison, see the best AI product photography tools for ecommerce, the best AI virtual try-on tools and the complete guide to AI product photography for e-commerce.
There are also things NotShot does not do today. There is no public API, and every image is rendered at 4K, so there is no cheaper low-resolution tier. If you need hundreds of low-resolution images a day through an API, a different tool is a better fit.
