AI-Powered Virtual Try-On Tools: What Works in 2026

AI-powered virtual try-on tools show you how clothes, glasses, makeup, or shoes might look on you before you buy. If you're a shopper, the easiest place to start today is Google: search for shoes, a top, bottoms, or a dress, tap "Try it on," and upload a full-length photo. If you're a brand, the choice is between AR try-on you embed in your own site or app (Perfect Corp, Snap) and generative try-on that platforms like Google already run on your listings, often without you doing anything.

One thing every tool has in common: they show you the look, not the fit. Google says so on its own help page. Size charts and reviews still decide whether something fits.

Three kinds of virtual try-on technology

1. AR face and body tracking. Your camera runs live, and the software tracks your face, feet, or wrist in real time. A 3D model of the product sits on top. This is the mature option for makeup, eyewear, shoes, and watches, because those products are rigid or sit on a surface that's easy to track.

2. Generative try-on from a single photo. You upload one photo of yourself. An image model generates a new picture of you wearing the garment, including how the fabric drapes, folds, and wrinkles. This is what's made clothing try-on practical, since clothes don't behave like rigid 3D objects. The output is a still image or a short video, not a live camera view.

3. Sizing from body data. This doesn't show you anything. It predicts which size to order, using your measurements, your past purchases, or what other shoppers said about fit.

Try-on is one of the clearer consumer examples in our breakdown of generative AI use cases.

Quick comparison: AI virtual try-on tools in 2026

The tools, one by one

Google Try-On in Search and Shopping

Per Google's Try-On help page, you search for a top, bottom, or dress, click "Try it on," and upload a photo. It works in 18 countries. You need to be 18 or older, signed in, and have Search history and personalized recommendations turned on.

Those are the examples the shopper-facing page walks you through, not the full list of what's covered. Google's Merchant Center documentation states the supported categories as shoes, tops, bottoms, and dresses, and excludes lingerie, bathing suits, and accessories. So if you're shopping for footwear, it's worth looking for the Try-On option even though the shopper walkthrough doesn't use it as an example.

For better results, Google recommends a head-to-toe photo or a selfie, with good lighting, a clean background, hands visible, and fitted clothing. In the US you can also create a digital version of yourself from a selfie, using Google's Nano Banana image model, and reuse it.

For brands, Google's Merchant Center help says any merchant with a shopping feed that's eligible for free listings is automatically opted into Try-On, across all four supported categories — shoes, tops, bottoms, and dresses. The feature isn't available for Shopping ads, and merchants can opt out.

Google Labs also ran Doppl, a separate experimental try-on app. Google shut Doppl down on April 30, 2026 and now points users to Try-On. If you mostly shop from your phone, our guide to using AI on your phone covers what else is worth setting up.

Amazon Virtual Try-On

Amazon runs both kinds of try-on. Its Virtual Try-On page says apparel try-on, on clothing product pages in the US, uses AI to render clothing onto an image of you, while eyewear try-on uses AR in the US and Canada. Earlier, Amazon launched Virtual Try-On for Shoes lets you point your phone at your feet and see sneakers from every angle, in the Amazon Shopping app on iOS, for customers in the US and Canada. Amazon also describes AR try-on for sunglasses, lip colors, and eyeshadow in its app.

On fit, Amazon leans on the third category: AI size recommendations and fit insights pulled from customer reviews.

Perfect Corp, for beauty and eyewear brands

Perfect Corp sells try-on to businesses. Its eyewear solution tracks your face in real time, measures pupillary distance automatically, and installs as a plugin on Shopify, Wix, and Squarespace. In January 2026 the company announced generative fashion APIs for watches, bracelets, rings, earrings, necklaces, scarves, hats, shoes, and bags, alongside its existing clothes and fabric APIs.

Snap Lens Studio and Camera Kit

Snap's try-on tooling for ecommerce includes a Clothing Try-On feature built on 3D body mesh and body tracking, plus templates for footwear, eyewear, wrist accessories, and makeup. You build the Lens in Lens Studio. Camera Kit then lets you embed it in your own iOS, Android, or web app, so shoppers don't need Snapchat.

IDM-VTON and open research

If you're technical and want to see how generative try-on works under the hood, IDM-VTON is a well-known diffusion model for virtual try-on, published at ECCV 2024, trained on the VITON-HD dataset and evaluated on VITON-HD and DressCode. Its license is CC BY-NC-SA 4.0, which doesn't allow commercial use, so it's for learning, not your store.

What virtual try-on gets wrong

Look is not fit. A generative model produces a believable image. It doesn't know your measurements or the garment's real ones. Google's help page states plainly that the visualizations don't determine or guarantee actual fit.

Body diversity is uneven. Google's own guidance asks for fitted clothing and a clear full-body pose, and warns that generated images can contain errors in body shape. Loose clothing or an unusual pose in your photo makes that more likely. Test with a range of real body types, not just the demo photo.

Fabrics and details slip. Google warns generated images can get clothing details wrong. Expect the most trouble with sheer or shiny fabrics, small prints, and logos. With AR makeup, room lighting changes how a shade reads, so check the swatch too.

Privacy of your body photos

A full-length photo of you is sensitive. Before uploading, check three things: whether it's used to train models, whether you can delete it, and who else sees it.

Google's help page says it doesn't collect biometric data from try-on photos, doesn't use them for training, doesn't share them with other Google services or third parties, and lets you delete them anytime. Hold smaller apps to that baseline, and skip any app that won't say. The same caution applies to other AI photo trends, like the Gemini Polaroid prompt.

How brands should evaluate a virtual try-on tool

Perfect Corp's announcement says try-on helps lower product return rates. Snap's ecommerce post says AR try-on reduces return rates. Neither page gives a number or cites a study. Treat those as the vendor's claim, and ask for the data behind it.

Google's 2023 try-on announcement cites a Google/Ipsos survey: 59% of US online clothing shoppers said they'd been dissatisfied with an item because it looked different on them than expected. That's a problem statement, not proof any tool fixes it. A sensible evaluation:

1. Match the tech to the product. AR for eyewear, beauty, shoes, and watches. Generative for clothing. Sizing tools if your returns are about fit.
2. Check what's already running. If you sell shoes, tops, bottoms, or dresses with a Google shopping feed, your products may already appear in Google Try-On.
3. Run a controlled test. Split traffic on a set of product pages, with and without try-on, and measure add-to-cart, conversion, and returns over a full return window.
4. Test on real bodies and hard fabrics. Put your trickiest products through it with photos from a range of body types, skin tones, and ages.
5. Read the data terms. Where are shopper photos stored, for how long, and does the vendor train on them?
6. Check the license. Open models like IDM-VTON can be non-commercial.

If you've found a try-on or product-photo workflow someone shared online but can't get it running, Taku lets you mirror an AI app or workflow someone already got working into your own desktop workspace, run it, and remix it for your catalog. Taku is in Beta, and the Mac app is available now.

Key Points

• AI-powered virtual try-on tools come in three types: AR tracking for makeup, eyewear, and shoes, generative try-on for clothing, and sizing tools for fit.
• Google Try-On is the widest-reaching clothing option, in 18 countries for shoes, tops, bottoms, and dresses. Google's separate Doppl app shut down in April 2026.
• Amazon runs generative apparel try-on in the US and AR try-on for eyewear, shoes, and beauty, with AI size recommendations for fit.
• Perfect Corp and Snap serve brands: Perfect Corp through AR and generative APIs, Snap through Lens Studio and Camera Kit.
• IDM-VTON is open research code under a non-commercial license.
• Every tool shows the look, not the fit. Return-rate reductions are vendor claims until you test them yourself.

FAQ

Can AI virtual try-on tools show how clothes will actually fit?

No. Generative try-on shows how a garment might look on you, not whether your size fits. Google says its visualizations don't guarantee fit and points shoppers to size charts and reviews.

How do I virtually try on clothes with AI on Google?

Search for a supported product on Google Search or Shopping — the covered categories are shoes, tops, bottoms, and dresses — click "Try it on," and upload a well-lit, full-length photo of yourself. You need to be 18 or older, signed in, and in one of the 18 supported countries.

Are virtual try-on apps safe for my photos?

It depends on the app. Look for a clear statement that photos aren't used for training, can be deleted, and aren't shared. Google publishes all three for Try-On. If an app won't say, don't upload.

What is the best virtual try-on tool for makeup and glasses?

AR, because faces track well in real time. Shoppers can use AR try-on inside retailer apps such as Amazon's. Brands usually license it from a vendor like Perfect Corp, or build Lenses with Snap.

Do virtual try-on tools reduce returns for ecommerce brands?

Vendors say so, but the pages we checked from Perfect Corp and Snap give no numbers. Treat it as a hypothesis and split-test it on your own product pages first.

Can I use open-source virtual try-on models like IDM-VTON in my store?

Not IDM-VTON. Its CC BY-NC-SA 4.0 license rules out commercial use. Check the license of any open model before using it commercially.