The short answer
Virtual try-on uses the phone camera to track a face, hand, foot or body and places a 3D product on it in real time. Eyewear, makeup and hats are mature and convincing. Watches, rings and shoes work well. Clothing is the hardest and still improving. A focused demo with a few products starts around $999 and a full try-on application around $2,999 with our AR packages. The larger cost for most retailers is turning a catalogue into good 3D models.
What works today
| Category | Tracking used | Maturity |
|---|---|---|
| Glasses and sunglasses | Face mesh | Excellent. Stable, realistic, widely adopted |
| Makeup and hair colour | Face and feature segmentation | Excellent for lips, eyes and foundation |
| Hats and headwear | Head pose | Very good |
| Earrings and necklaces | Face and neck landmarks | Good |
| Watches and bracelets | Wrist and hand tracking | Good, with care over occlusion |
| Rings and nails | Finger tracking | Good, small scale is demanding |
| Shoes | Foot tracking | Good from typical angles |
| Clothing | Body tracking and cloth simulation | Hardest. Fit and drape remain approximate |
Furniture and homeware are a different technique, placing objects in a room, which we cover in AR furniture app development.
How the tracking works
Phones detect a face and fit a 3D mesh to it many times per second. Apple's ARKit uses the depth camera on supported iPhones for precise face geometry, and Google's ARCore Augmented Faces provides a face mesh on a wide range of Android devices using the ordinary camera. Machine learning models do the same for hands, feet and whole bodies. The product model is attached to the tracked points and rendered with lighting that matches the scene.
Two details separate good try-on from poor:
- Occlusion. The arms of glasses must disappear behind the ears, and a watch strap behind the wrist. Without it the product looks pasted on.
- Scale. Shoppers want to know if a frame is too wide. Accurate sizing needs depth data or a calibration step.
3D assets are the real project
Every product needs a 3D model that is accurate, light enough to load quickly and has realistic materials. Glass, metal, gemstones and fabric are demanding. Options:
- Modelling from photos and measurements. Flexible and common. Price depends on complexity.
- Converting design files. If products are designed in 3D, models can be optimised from those files.
- Scanning. Photogrammetry works for matte objects and poorly for shiny or transparent ones.
- Variants. One model with swappable colours and materials covers many items cheaply.
For a catalogue of hundreds of products, asset production usually costs more than the software. Set up a pipeline with templates, naming rules and quality checks from the start. See 3D modelling cost for how prices scale with detail.
Where to deliver it
| Option | Strength | Limit |
|---|---|---|
| WebAR on the product page | No install, direct path to purchase | Tracking and rendering lighter than native |
| Inside your own app | Best quality, device depth sensors, saved looks | Shoppers must have the app |
| SDK from a try-on vendor | Fast to launch, proven tracking | Licence fees, less control |
| Social camera lenses | Reach and sharing | Purchase path is indirect |
Try-on on the product page, in the browser, usually has the most effect on sales because it sits where the buying decision happens. We compare the technologies in ARKit vs ARCore vs WebXR.
Build or license
Vendors sell ready try-on engines for eyewear, cosmetics and footwear, charged by subscription or usage. They are quick to adopt and good for standard categories. A custom build makes sense when your product does not fit a standard category, when you want to own the experience and data, when licence fees at your scale exceed development cost, or when try-on is part of a larger app. Many projects combine both: licensed tracking with a custom interface and catalogue integration.
What it costs
| Scope | Typical cost |
|---|---|
| Demo with 3 to 5 products in one category | From $999 |
| Full try-on application with catalogue and capture | From $2,999 |
| Store integration, analytics and 50 to 200 products | $20,000 to $60,000 |
| Multi-category, custom tracking, large catalogue | $60,000 to $100,000+ |
| 3D model per product | Tens to a few hundred dollars depending on complexity |
Drivers are the category, the number of products, the realism of materials, platforms and the depth of integration with your store. See AR app development cost.
Features shoppers use
- Photo and video capture with sharing.
- Side-by-side comparison of two products.
- Size or fit recommendation.
- Add to basket without leaving the camera view.
- Saved favourites.
Privacy
Try-on processes camera images of faces and bodies. Process on the device where possible, do not store images without clear consent and explain plainly what is collected. Biometric data laws in some regions, including parts of the United States and the European Union, impose specific duties, so involve legal advice early.
Measuring results
Compare shoppers who use try-on with those who do not: conversion rate, average order value, time on page and return rate. Run it as a controlled test where possible, since people who open try-on may already be more interested. Our overview of AR in retail discusses what to expect.
Common mistakes
- Launching with poor 3D models. Bad models reduce trust in the product itself.
- Hiding the feature behind several taps.
- No fallback for unsupported devices.
- Ignoring load time on mobile networks.
- Promising exact fit for clothing.
Getting started
Pick your best-selling category and five products. Send us product photos, dimensions or 3D files, and tell us whether you want web or app delivery. We will build a demo at a fixed price so you can test it with real shoppers. Packages are on the VR and AR offer page.
Sources: Google ARCore: Augmented Faces · Apple: Augmented reality · Wikipedia: Virtual dressing room. Checked November 11, 2026.
Related: AR game development · AR furniture app development · Get a quote
Quick answers ❓
What is AR virtual try-on?
It is a feature that uses a phone camera to track the face, hands, feet or body and show a 3D product on the shopper in real time.
Which products work best with virtual try-on?
Glasses, makeup and hats work best. Watches, rings and shoes work well. Clothing is the hardest because fit and drape are difficult to simulate.
How much does a virtual try-on app cost?
A demo with a few products starts around $999 and a full application around $2,999. Store integration with larger catalogues costs $20,000 to $100,000 or more.
Does virtual try-on need an app?
No. It can run in the browser on the product page, inside your own app, through a vendor SDK or as a social camera lens.
What is the biggest cost in a try-on project?
Usually creating accurate, lightweight 3D models for every product in the catalogue.
Related guides
AR in retail → AR app development cost →Want virtual try-on for your products? 🕶️
We build AR try-on for web and apps. Demo from $999.