Upload a photo of a person and one or more garment photos, and get back the same person wearing the new outfit, face, pose and body shape untouched. Mix tops, bottoms and accessories in one pass, compare looks in minutes, and skip the reshoot until you already know the outfit works.
An outfit looks different on every body, and no product photo on a rack can tell you how it will look on yours, or on the model you booked for tomorrow. Virtual try-on closes that gap without a fitting room or a studio day: upload a photo of the person and a photo of the garment, and get back the same person wearing it, in about the time it takes to read this sentence.
A flat-lay photo of a jacket tells you the color and the cut. It does not tell you how the shoulders sit on a real body, whether the length works with the rest of an outfit, or how a pattern reads against a specific skin tone. Shoppers guess anyway, and a share of them guess wrong, which is most of why clothing has the highest return rate of anything sold online. Sellers feel the other side of the same problem: photographing one garment on five models in six colorways is a full shoot, priced and scheduled like one.
Virtual try-on models close that gap by editing a real photo instead of generating a new one from scratch. The person's face, pose and body stay put; only the clothing changes. That single constraint is what makes the output useful rather than merely interesting, because a shopper is judging their own body in the outfit, not a stranger's.
Picasso IA runs 488 models behind one interface, and a dedicated garment-swap model sits inside that catalog alongside general-purpose editors you can reach for the same task. New accounts start with free credits, enough to run several outfits before spending anything, and the full model catalog is worth a browse if a specific look calls for a different tool.
Most image editors can be talked into a clothing swap with a careful prompt. P Image Try On, from prunaai, is built for exactly this job: a person photo plus one or more garment photos, up to six recommended and eleven supported in a single run, mixing a top, a bottom, shoes and an accessory in one pass instead of one edit per item.
The output keeps the person's face, body shape and pose exactly as they were, at the same pixel resolution you uploaded, so it drops straight into a listing or a moodboard without a resize step. Turbo mode speeds up runs with four or fewer garments, which covers most everyday outfits. For a garment photo that shows more than one item, an experimental prompt field lets you point at the piece you actually want, for example the green jacket from the second photo, so a busy lookbook shot does not force you to crop first.
For simpler single-item swaps, or a more stylized edit than a straight placement, general editors in the catalog do the job too. Nano banana and Seedream both take a photo plus a plain instruction and suit a smaller change, say swapping a shirt color or adding a scarf. FLUX Kontext handles targeted edits well when only one detail should move. P Image Try On stays the better default whenever more than one garment needs to land together.
What you upload changes what you get back. A quick reference for planning the shot before you generate anything:
| Look You Want | Garments to Upload | Turbo Mode | Best For |
|---|---|---|---|
| Single item swap | 1 garment photo | Turn on | Testing one piece fast |
| Full outfit | Top, bottom, shoes | Turn on | Catalog and listing shots |
| Layered look | Jacket, top, bottom, accessory | Turn on | Editorial and lookbook styling |
| Head-to-toe with accessories | 5 to 6 garment photos | Turn off | Campaign and hero images |
| New pose, then dressed | 1 to 3 garments plus a reference pose | Turn off | Consistent poses across a catalog |
Turbo mode is meant for four garments or fewer; past that, running the full-quality pass gives the model more room to place each piece correctly.
This runs end to end inside the Picasso IA toolkit and takes a few minutes once the photos are ready. It assumes a clear photo of the person and one or more garment photos, ideally flat-lay or on a plain background.
The gap between a garment on a hanger and a garment on a body is mostly pose and proportion, and that is exactly what a preservation-based model holds onto. Because the source photo's pose carries through, the result reads as the same person in a different outfit, which is what makes it useful for a real purchase decision or a client approval, not just a nice picture. A reference pose image pushes this further: it repositions the person before garments are applied, useful when the original photo is seated but the shot needs to be standing.
Upload the person and garment photos at similar lighting and color temperature. A daylight photo paired with a garment shot lit under warm indoor bulbs asks the model to reconcile two different scenes at once, and that compromise shows up as slightly off color matching on the finished outfit.
For a shopper, the use is direct: hold a piece you are considering against your own photo before you order it, and skip the return that starts with it looked different in person. E-commerce sellers get more at scale, one product photographed once and shown across colorways and body types without rebooking a model for every variant, which pairs naturally with AI product photography for the surrounding shots. Stylists use it for planning rather than final delivery, mocking up outfit directions before anyone packs a garment bag. Teams also managing on-camera talent lean on the same reference-image discipline used for consistent AI characters to keep identity steady while outfits change. If your team is weighing this against a dedicated product-photo editor, Picasso IA vs Photoroom covers where each pulls ahead.
An honest description of the tool includes where it stops working. For outfit try-on, five places to expect trouble:
None of this makes the result unusable; it makes it a very strong first pass that still benefits from a look before it goes to print or to a live listing.
Picasso IA gives new accounts free credits, and a garment swap is a single image generation, which sits toward the cheap end of what the catalog charges for. That is enough to test several outfit combinations before spending anything. Once the free credits run out, current plans are listed on the pricing page, since costs change often enough that a number written here would go stale.
Yes. P Image Try On accepts up to six garment images recommended, and up to eleven supported, in a single run, so a top, a bottom, shoes and an accessory can all land in one generation instead of one edit per piece. Turbo mode is meant for four garments or fewer; beyond that, running without turbo gives the model more room to place every item correctly.
It works best with a clear, front-facing photo in even light against a plain or simple background. Photos with an unusual angle, heavy shadow or a pose where a limb crosses the body are harder for the model to work with, and results get less reliable the further the source photo sits from a natural standing pose. A phone photo taken in daylight is usually enough.
P Image Try On, from prunaai, is purpose-built for this task. It takes a person photo and one or more garment photos and returns the same person dressed in the new outfit, with the original face, pose and body shape kept intact, at the same resolution as your source photo. General editors like nano banana and FLUX Kontext can handle a smaller single-item swap, but P Image Try On is the model to reach for whenever more than one garment needs to land together.
Yes, and that is one of the more common uses. A single model shoot can be extended across colorways and multiple garments without rebooking a photographer for every variant, which is useful for sellers testing a new option before committing to inventory. Pair it with AI product photography for backgrounds and lighting once the outfit itself looks right.
Yes, that is the point of a preservation-based model rather than a generic one: your body shape and pose from the source photo carry through, and only the clothing changes. What can shift slightly is how a fabric behaves, since the model renders how a garment appears to sit rather than simulating its actual weight, which matters more for loose or heavy pieces than fitted ones.
Yes, with a reference pose image. Supply a second reference showing the pose you want, and the model repositions the person to match it before the garments are applied. This is useful when your source photo is seated but the outfit needs to be shown standing, though it is marked experimental and can occasionally fail on a difficult seed, so it is worth a retry if the first result looks off.
Use the experimental prompt field to describe the exact piece you want, for example the blue jacket from the second photo, and the model uses that instruction to pick it out of a busier image such as a lookbook shot. This saves cropping the source photo yourself and works well for garment photography that was never meant to isolate a single item.
The model returns output at the same resolution as your uploaded person photo, so a high-resolution source already helps. If it still needs to be larger, an upscaling model such as Clarity Pro Upscaler enlarges it without the softness a simple resize introduces, which matters for a large listing image or a printed piece.
Not entirely, and it is not meant to. It is strongest as a fast way to test which outfit or combination is worth the time and budget of an actual shoot, and as a preview a shopper or a client can react to before anything ships or gets booked. For a final campaign image where fabric weight and exact texture must be right, a real shoot still earns its place; virtual try-on decides which looks make the cut.
Stop guessing how an outfit will look and see it on the real photo. Open the Picasso IA toolkit and run your first garment swap in the next few minutes.
