Virtual try-on vs AI model photos: what’s the difference?
Virtual try-on shows a garment on the shopper’s own photo, or on an avatar made from it, while they shop. AI model photos are catalogue and campaign images a brand makes in advance and publishes like any product photo. Try-on answers how something might look on one person; model photos show every shopper the same approved view. Both redraw the garment, and neither measures fit. Google’s own help page says its try-on does not guarantee how a garment fits.
The difference in one table
| Question | Virtual try-on | AI model photos |
|---|---|---|
| Who makes the image | The shopper, from their own photo or avatar, while shopping | The brand, before the product goes live |
| Who sees it | Only that shopper | Every shopper, on product pages, marketplaces and ads |
| Who checks it before anyone sees it | Nobody | An art director, who can approve or reject each photo |
| The question it answers | How might this look on me | What does this look like worn |
| Where it runs | Google Search and Shopping, retailer apps such as Zara’s, or a try-on widget on your own store | Your store, marketplaces, ads, email and lookbooks |
| Cost | Free for shoppers on Google; about $0.06 an image if you build your own on Google Cloud | About 3 to 24 cents per render on Google’s image models, plus review time |
How Google’s try-on works
Google opened its current try-on in Search Labs, its opt-in testing program, in the US on May 20, 2025. A shopper uploads a full-length photo and sees shirts, pants, skirts and dresses from Google’s shopping listings on their own body. Google said it built an image model for fashion that understands how materials fold, stretch and drape on different bodies. It added shoes on October 8, 2025, and in December 2025 let US shoppers make a full-body version of themselves from a selfie.
As of September 2026, Google’s help page lists 18 countries, including the US, UK, Canada, Australia and Japan. It says lingerie, swimwear and accessories are not supported, and that the "Try it on" button does not appear on sponsored Shopping ads. Google says it does not keep biometric data from the photo or use it for training.
For merchants, Google’s Merchant Center help says brands with a shopping feed and high-quality images are opted in automatically. It wants one garment on one front-facing model or mannequin in a simple pose, or laid flat, at least 512 x 512 pixels and ideally 1,024 or more, with no hands or bags covering details. Google says it shows only model sizes that match the sizes a brand sells, and merchants can opt out of try-on in paid ads through Merchant Center support.
Brand-side try-on: Zara, and building your own
Zara runs its own. Inditex says Zara Try-on, live since mid-December 2025, lets customers build a synthetic avatar from their own photos and see it wearing real products. In March 2026 it reported the tool in 43 markets with more than 7 million sessions, on Zara.com only for now.
Smaller brands can build try-on from APIs, services a developer can call from your store. Google Cloud’s Vertex AI lists its Virtual Try-On model at $0.06 per image; it takes one photo of a person and one of the garment and returns one to four images. FASHN, a fashion AI company, sells plans from $19 a month and charges most tools at one credit per output image. Either way, you pay for every image a shopper makes, and no one on your team sees it first.
What neither one can promise
Both redraw your garment with an image model, so both can change details, and neither measures cloth against a body. Google’s help page calls the try-on image an approximation that may not reflect body shape, personal features or product details accurately, and sends shoppers to size charts and reviews for fit. Doppl, Google’s experimental app that turns try-ons into short videos, carries a similar warning about fit, appearance and clothing details.
With AI model photos you can catch a bad image before it goes live, because a person reviews each one. With try-on, the check has to happen before the shopper arrives: clean product images, accurate garment data and an honest size chart. Put the model’s size under your photos too; see how to show clothes on different body sizes.
When to use which
| Situation | Use | Why |
|---|---|---|
| You need product page and ad images for a new drop | AI model photos | Every shopper sees the same approved images on launch day |
| Shoppers hesitate over whether a style suits them | Try-on, next to model photos | It shows the garment on them, which a model photo cannot |
| You sell on Google Shopping in a try-on country | Clean flat lay or front-facing model images | Google may already use them, and image quality decides eligibility |
| Fit is your main reason for returns | Neither on its own | Size charts, fit notes and model measurements carry fit information; images do not |
| You sell swimwear, lingerie or accessories | AI model photos or a shoot | Google’s try-on does not support these categories |
To make the model photos, see how to turn a flat lay into on-model photos and AI fashion photography.
Questions people also ask
- Does virtual try-on reduce returns?
- We did not find a published, independent study of return rates before and after try-on that we could verify. Treat vendor claims as claims, and compare your own return rates by product before and after you add it.
- Do I need my own try-on if Google already has one?
- Google’s try-on appears only on Google, for supported categories, in 18 countries. If most of your traffic goes to your own store or app, a try-on there reaches shoppers Google’s does not, at a per-image cost that you pay.
- Will try-on replace model photos?
- Not in the tools we checked. Google’s try-on starts from your product images and asks for garments on a front-facing model or laid flat, so it still depends on the catalogue photos you make.
Where Overs fits
Overs makes the second kind: catalogue and campaign photos made before launch from your product photo, with one approved model across the set and a review step where you approve, change or reject each photo.
Free for 40 photos a month. The AI that makes the photos is billed separately, on your own key, with no markup from Overs.
Sources
- Google: Shopping on Google, AI Mode and virtual try-on updates from I/O 2025 (May 20, 2025)
- Google: Google Shopping’s try on tool adds shoes, expands to new countries (October 8, 2025)
- Google: Virtual try on adds studio-quality images (December 11, 2025)
- Google Shopping Help: How the Google try-on tool works
- Google Merchant Center Help: About apparel virtual try-on
- Google: Doppl, a Google Labs app (June 26, 2025)
- Inditex: FY2025 results (Zara Try-on)
- Google Cloud: Vertex AI generative AI pricing (Virtual Try-On)
- Google Cloud: Generate Virtual Try-On images
- FASHN: Pricing
- Google AI for Developers: Gemini API pricing