Dev.to WebDev 🛠 Dev 👁 0 📖 3 min read

Virtual Try-On in 2026: A Map of the Landscape, and 5 Design Decisions Behind Every Product

Disclosure: we build Pastely, one of the products mentioned below. We have tried to describe every tool, including ours, by what it is built for. Virtual try-on has had a busy year. Doji raised $14M in 2025 to make avat

Virtual Try-On in 2026: A Map of the Landscape, and 5 Design Decisions Behind Every Product

Disclosure: we build Pastely, one of the products mentioned below. We have tried to describe every tool, including ours, by what it is built for.

Virtual try-on has had a busy year. Doji raised $14M in 2025 to make avatar try-ons fun, Google shut down its standalone Doppl app in April 2026 and moved try-on into Search, and ASOS launched virtual try-on in February 2026 on around 10,000 products in its iOS app.

If you are building anything in this space, or are just curious how these products differ, here is the map we use.

Three product shapes

1. Shopper apps. The user tries clothes on themselves before buying or getting dressed.

  • Avatar or likeness apps: Doji (iPhone, iOS 18+) builds an AI likeness from a few photos and adds a style feed and a browser extension.
  • Closet apps: Alta (iPhone, Android, web) and Whering start from the clothes you already own.
  • Outfit editors: Pastely (web) lets you combine pieces into one outfit and preview it on your photo.

2. Creator tools. The goal is an image to post, not a purchase decision. Magic Hour swaps one garment onto a person photo; Dreamina turns an outfit idea into a styled visual.

3. Commerce tools and APIs. Sellers need model shots or try-on at scale: FASHN AI (with an API) and Photoroom's virtual model, plus try-on built into retail surfaces like Google Shopping listings and ASOS.

Five design decisions that shape the product

These are the choices we had to make, and every product in the list above made them differently.

Platform: native app or web. A native iPhone app can feel more polished, but it excludes Android and laptop users. We went web-first so the same editor works on any device without an install.

Unit of try-on: one garment or a whole outfit. Trying on a single product is the simplest flow and fits a product page. But the question people actually ask is "does this jacket work with these trousers and these shoes?", so we made the outfit the unit.

Identity input: likeness, single photo, or sample model. A likeness built from several photos gives consistent results but adds setup friction. A single photo is faster. A sample model lets people start without uploading anything, which matters for privacy-conscious users. ASOS chose a hybrid: shoppers either upload a photo or pick an AI model that resembles them.

Context: one image or several settings. The same outfit reads differently in a studio, on a street and in a park. Showing several settings helps people judge the look, not just the garment.

The same outfit previewed in three settings: studio, street and park

Honesty about fit. A preview shows how a look reads: colour, proportion, style. It does not tell you how a garment fits. Every try-on product should say so clearly, and point to the size chart.

What a good input photo looks like

Whatever the model behind a try-on, it has to infer body shape, pose and lighting from one image. Results get much better with:

  • a full-body photo, facing the camera
  • even, natural light
  • fitted clothes, so the body outline is visible
  • a plain background

If your product takes user photos, putting these four tips right next to the upload button is the cheapest quality improvement you can make.

Try it, and tell us what breaks

You can try Pastely in the browser: open the editor, pick a top, bottoms and shoes, and preview the outfit on your photo or a sample model. There are free daily previews and no account is needed to start.

We are launching on Product Hunt on November 24, and honest feedback, especially from developers, is very welcome in the comments here.

Sources: TechCrunch on Doji, Doji on the App Store, Google Labs (Doppl), ASOS, Alta, Magic Hour, Dreamina, WearView. Facts checked in November 2026.

📰 Read the original article on Dev.to WebDev

Originally published by Dev.to WebDev. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.