Which fashion brands use AI virtual try-on? Zara and Zalando

A five-looks-a-day cap is not a demo limit. It is a company managing real generation cost and real server load for a feature already live to shoppers. That detail alone says more about where this technology stands than any press release would.

Which fashion brands are already using AI virtual try-on?

Bloomberg's August 2026 reporting confirmed both Zara and Zalando have shipped their own virtual try-on features, built by two different teams solving the same problem in two different ways. The article opens with the reason either company bothered: shoppers routinely order several sizes of the same item with a plan to keep one and return the rest, a habit that costs the industry money in shipping, restocking, and processing on garments that were never going to stay sold.

• Zalando built a fit-prediction system around body measurements paired with 3D scans of the actual garments in its catalog.
• Zara built a generative system: a shopper uploads a photo, and the tool renders them wearing a chosen outfit on an avatar meant to resemble them.
• Neither is a side project. Both are live, shopper-facing features inside apps used by hundreds of millions of people, not a pilot running in one test market.

How does Zalando's approach differ from Zara's?

Zalando's method starts from measurement, not imagery. It compares a shopper's own body data against a 3D scan of the garment and reports back on fit, closer to a very detailed size chart than a picture. Zara's method starts from a picture: the output is an image of a person wearing an outfit, generated fresh for that shopper and that item.

Bloomberg reports Zalando is now working to combine the two: its own precise, measurement-based fit prediction layered under Zara's more visual, generative style. That a company already running one approach at scale wants the other one too is worth sitting with. Neither method alone was judged good enough on its own.

What are the limits of these in-house tools?

The same Bloomberg piece tested Zara's tool directly and found real friction, not just a rough edge or two.

• Generation speed. Each outfit took about two minutes to render, per the reporter's own use of the tool.
• A daily cap. The tester was limited to five generated looks per day, a real usage limit on a live consumer feature.
• Comfort with avatars. Zalando's own research found that seeing a virtual representation of your own body can lower a shopper's confidence rather than raise it, not a universal win.
• Feel is still missing. Neither tool can show how a fabric actually drapes or stretches on skin. A render answers the fit question, not the texture question.

What does this mean for a smaller fashion store?

Two of the largest fashion retailers in the world spending real engineering time on this is a signal worth reading correctly. It does not mean a smaller store needs a 3D scanning rig for every garment, or a team that builds and maintains its own generative model. It means the underlying idea, showing a shopper what an item looks like on them before they buy, has moved from a nice-to-have experiment to something two of the biggest names in the industry now treat as core infrastructure.

The part that does not scale down is the build. Zalando's 3D garment scanning alone is a catalog-wide operational project. Zara's generative pipeline needed a team to build, host, and moderate an image model at the volume of one of the largest fashion retailers on earth. A smaller store gets the same underlying idea, a photo in and a rendered outfit out, through a hosted tool instead of an internal engineering project.

How does Corlen compare to building an in-house tool?

Corlen follows the same basic shape as Zara's approach: a shopper's photo and a garment photo go in, a rendered result comes out, closer to Zara's generative style than Zalando's measurement-only one. Under the hood it runs on Gemini 3 Pro Image, also called Nano Banana Pro, the same image model family behind Perplexity's own Virtual Try On feature, covered in more detail in what AI shopping search means for your store. The difference from Zara and Zalando's tools is where the work happens: neither of their systems runs on anyone else's product pages, while Corlen installs directly onto a store's own Shopify or WooCommerce product page, or through the developer API for a custom build.

See it directly first: try Corlen on your own photo, no account needed for a first result.