ZURICH/BARCELONA / AGILITYPR.NEWS / May 28, 2026 / E-commerce has never solved the fitting room problem. You can zoom, rotate, read reviews, and

check the size chart. But you can’t try things on. For made-to-measure clothing, where every

garment is unique, the problem is even worse: there’s no sample to try, no display model to

reference. Every order is a one-off.

Hockerty, the online platform for custom menswear, just solved it with a Virtual Try-On powered

by a multi-layered AI pipeline that goes far beyond simple image overlays.

The try-on doesn’t sell the garment. It answers a question the customer already has.

How it works

The system runs a multi-step process. First, it analyzes the garment the customer has

configured: fabric, cut, colour, lapel style, pocket type, every detail. Then it determines the

optimal way to render that specific combination. Finally, it generates an image of the customer

wearing the garment as if it were made to their exact measurements.

This is not an overlay. And it’s not a filter. The AI makes intelligent decisions about how to display

each garment type. A double-breasted blazer renders differently from a slim-fit shirt. A peak lapel

catches light differently from a notch lapel. The system accounts for all of it.

The main engineering priority was garment consistency: ensuring the rendered output matches

the exact configuration the customer selected. Hockerty has achieved this across its full catalog,

from suits to jeans to sneakers.

At the time of release, the system generates a try-on image in 20 to 30 seconds. The team

expects this to improve as the underlying models evolve.

The technology is not limited to suits. It works across Hockerty’s entire range: jeans, field

jackets, polos, overcoats, sneakers, dress shoes. Any garment a customer designs in the 3D

configurator can be rendered onto their photo.

“We started with one AI model. We’ve already swapped it three times. The pace of improvement

in generative AI means we’re not building on a fixed technology. We’re building on a moving one.

And that’s actually the point.”

Alberto Gil, Co-Founder, Hockerty

The infrastructure evolves monthly

Hockerty’s Virtual Try-On was built on specific AI image generation models, but the underlying

technology changes almost monthly. As faster, more reliable models emerge, the team integrates

them. The model powering the try-on at the time of this publication may differ from the one in use

a month later.

This approach treats the AI layer as infrastructure, not product. The product is the experience:

upload a photo, see yourself in your suit. The technology behind it is deliberately fluid.

An optional step further

Customers can also place themselves in a different environment: a boardroom, a wedding

venue, a city street. This runs an additional rendering process. Current limitation: the

environment swap can slightly alter facial expressions. The team is actively working to resolve

this without compromising speed or usability.

“We could have waited until everything was perfect. But in AI, waiting means falling behind. We

launched what works today and we improve it every week.”

Alberto Gil, Co-Founder, Hockerty

Same input. Different output. Same precision

The simplest way to understand what the system does: take one customer photo and render two

completely different garments. A beige linen suit. Then a green tuxedo with satin lapels. But the

same logic applies to jeans and a field jacket, or a polo and sneakers. The AI handles all of

them. Every button, every fold, every shadow recalculated from scratch. Not a filter swap. A full

re-render.

Not just one piece. A complete look

Whatever a customer designs in Hockerty’s 3D configurator, the Virtual Try-On renders it on

them. A linen suit for a wedding. Jeans and a field jacket for the weekend. A polo and sneakers

for after work. An overcoat for the morning commute.

Why this matters beyond fashion

Hockerty’s approach demonstrates a pattern likely to spread across e-commerce: using

generative AI not to create marketing content, but to solve a functional problem. That distinction

matters as the industry figures out where AI adds value and where it’s just noise.