When it comes to ar virtual try on, virtual try-on promises something simple: see the product on yourself before you buy it. Glasses on your face, a watch on your wrist, a lipstick shade on your lips, a sofa-scale check that a jacket actually suits your frame. The promise is real, the technology is genuinely impressive — and the gap between what it does brilliantly and what it still struggles with is exactly what retailers need to understand before investing.
This is a deep dive into how virtual try-on works, where it excels, where it falls short, and how to implement it without learning expensive lessons first-hand.
How Virtual Try-On Actually Works – Ar Virtual Try On
Every virtual try-on system does three things: track part of the body in real time, render the product onto that tracked region, and keep it stable as the person moves. The details vary enormously by product type.
Face tracking: eyewear, cosmetics, hats
The most mature category. Modern face tracking maps dozens of facial landmarks — the bridge and temples for glasses, lip boundaries for lipstick, eye regions for eyeshadow — and renders products onto them at 30 to 60 frames per second. Eyewear try-on is the gold standard here: glasses are rigid, faces are well-understood by tracking models, and the result is genuinely useful for judging style and proportion. Cosmetics try-on handles colour and finish well, though texture and true-to-life shade accuracy depend heavily on lighting and screen calibration. The same face-tracking technology also powers branded AR beauty filters on social platforms, which double as a product discovery channel.
Body tracking: apparel and footwear
The hard category. Full-body pose estimation tracks joints and segments, then drapes a 2D or 3D garment representation over the video feed. Footwear try-on — tracking the foot and rendering a shoe — works reasonably well because feet are relatively rigid and the camera angle is constrained. Full garments are the frontier: cloth drapes, folds, stretches and occludes in ways that are computationally expensive to simulate in real time on a phone. Current systems range from "surprisingly decent approximation" to "uncanny sticker on a torso," and honesty about which you're getting matters.
Hand and wrist tracking: watches, rings, jewellery
A sweet spot. Hands are complex but the interaction is constrained — the shopper holds their hand up and looks at one item. Watches and rings render convincingly because they're rigid objects on a relatively stable surface. Jewellery try-on converts well because these are high-consideration purchases where seeing scale on your own hand resolves the main hesitation.
What Virtual Try-On Does Well Today
Be specific about the wins, because they're real:
- Eyewear: style, proportion and fit judgement that genuinely reduces the need to order five pairs "to try at home."
- Cosmetics: shade exploration — trying twenty lipstick colours in a minute, which no physical counter allows. Shade accuracy varies; shade range exploration is excellent.
- Watches and jewellery: scale and presence on the body, the key uncertainty for high-value small items.
- Footwear: style and rough proportion checks, increasingly solid.
- Headwear and hats: straightforward tracking, reliable results.
The common thread: rigid or semi-rigid products, where the question is "how does this look on me" rather than "how does this fabric behave." For those products, try-on meaningfully improves the retail experience — shoppers explore more variants, decide faster, and return fewer items bought purely on guesswork.
There's a useful distinction to keep in mind: style try-on (does this suit me?) versus fit try-on (will this fit my measurements?). Style try-on is largely solved for the categories above. True fit prediction — this jacket in size M on your specific body — remains an active research area, not a solved product. Vendors sometimes blur this line; retailers shouldn't.
The Hard Parts Nobody Puts in the Sales Deck
Cloth simulation is genuinely difficult
A garment that looks right standing still and wrong the moment you move destroys trust faster than no try-on at all. Real-time cloth physics on mobile hardware forces compromises, and the compromises show on loose, flowing or layered garments. If your catalogue is mostly structured tailoring or knitwear, current tech copes. If it's draped eveningwear, set expectations accordingly.
Lighting and camera quality vary wildly
Try-on renders the product under the phone's current lighting, not studio lighting. A shade that looks perfect in the demo room can look wrong in a dim bedroom — and the customer blames the product, not their lightbulb. Good implementations include lighting estimation and honest on-screen guidance ("try in natural light for best colour accuracy").
Bodies are diverse; training data often isn't
Tracking models perform best on the body types most represented in their training data. Retailers serving diverse customers should test try-on across skin tones, body shapes, ages and assistive devices — wheelchairs, prosthetics — before launch. A feature that works beautifully for some customers and fails for others is a brand risk, not just a tech limitation.
Sizing is a data problem, not an AR problem
AR can show you a garment; it can't tell you your size unless the retailer has solved size data — consistent measurements, fit notes per garment, customer measurement profiles. Many "try-on" disappointments are actually sizing-data failures wearing an AR costume. Fix the size guides and fit data first; they're cheaper than the AR and they improve every channel.
The Business Case, Honestly Stated
Virtual try-on earns its keep through three mechanisms. First, confidence: shoppers who've seen a product on themselves hesitate less, which shows up in conversion on high-consideration items. Second, exploration: trying twenty shades or styles in a minute increases the variants considered per session. Third, returns: items bought after try-on are bought with better expectations, and better expectations mean fewer "not what I expected" returns.
What it doesn't do: it doesn't fix a bad product, a confusing site, or weak traffic. It's a multiplier on existing demand, not a source of it. Retailers sometimes expect try-on to rescue underperforming categories; it won't. It makes good categories perform better.
The economics hinge on catalogue shape. Eyewear retailers with a few hundred SKUs can cover the range with a contained 3D asset programme. Fast-fashion retailers with thousands of weekly new lines cannot — the modelling cost per garment kills it. Match the technology to a catalogue it can actually cover; our AR product visualisation service is built around getting this asset economics right.
Implementation: Build, Buy, or Blend
Off-the-shelf try-on platforms
Specialised vendors offer try-on as a service for eyewear, beauty, footwear and jewellery, typically as an embeddable web component or SDK. The advantages are speed and category-tuned tracking; the trade-offs are subscription cost, limited customisation, and dependence on a vendor's roadmap. For a first deployment in a standard category, buying is usually the right call.
Custom builds
A custom try-on experience — built with WebAR, native AR frameworks, or game-engine rendering — makes sense when the product or brand experience is non-standard: unusual products, signature interactions, or try-on woven into a larger AR retail experience. It costs more and takes longer, but you own the roadmap and the data. This is the work Visuosofts does — including dedicated try-on builds like our AR try-on service line.
The blend most retailers actually need
Buy the tracking, build the experience around it. Use a proven SDK for the hard computer-vision problem, but design the surrounding journey — entry points, product data integration, variant handling, analytics — as your own. The tracking is rarely the differentiator; the journey is.
WebAR versus native app
WebAR (browser-based, no download) wins on reach: shoppers try products from a link, an ad or a QR code with zero friction. Native apps win on performance and features — better tracking, saved profiles, offline use. For acquisition and first purchase, WebAR. For loyalty and repeat try-on within an existing app audience, native. Many retailers run both, with WebAR as the front door.
A Sensible Rollout Plan
- Start with your most try-on-suitable category — rigid products, high consideration, meaningful return rates. Prove the mechanism where physics is on your side.
- Pilot on WebAR to remove the app-download barrier, with entry points on product pages, category pages and paid social.
- Instrument honestly: try-on starts, completions, variant exploration depth, and downstream conversion and returns versus a control group. Vanity metrics like "filter opens" prove nothing.
- Test across devices and demographics before scaling — mid-range Android phones and diverse users, not just the team's iPhones.
- Expand the asset pipeline only after the pilot moves a commercial metric. 3D modelling is the recurring cost; scale it on evidence.
Virtual try-on is one of the few AR applications with a clear, repeatable commercial logic — for the right products. The retailers who benefit most are the ones who match the technology to their catalogue honestly, start with a contained pilot, and measure like sceptics. If that sounds like the approach you'd take, speak to our team for a free consultation — we'll tell you straight whether try-on fits your range, including when the answer is no.
Frequently Asked Questions
How accurate is AR virtual try-on?
It depends on the product. Eyewear, watches, jewellery and hats are highly reliable. Cosmetics are good for exploring shades, less exact for true colour. Full-garment try-on for soft clothing is still approximate — useful for style, not a substitute for sizing data.
Do customers need to download an app?
No. WebAR try-on runs directly in the mobile browser from a link or QR code. Native apps offer better performance and saved profiles, but WebAR removes the biggest adoption barrier for first-time users.
What's the difference between style try-on and fit prediction?
Style try-on shows how a product looks on you; fit prediction estimates whether a specific size will fit your measurements. Style try-on is mature for many categories; true fit prediction remains difficult and depends heavily on the retailer's size data quality.
How much does virtual try-on cost to implement?
Costs vary widely: off-the-shelf SDKs charge subscriptions, while custom builds are scoped projects. The largest ongoing cost is usually 3D asset production per SKU, which is why catalogue shape matters more than the technology choice.
Does virtual try-on reduce returns?
It can, for the right products — items bought after try-on are bought with better expectations. But it's a multiplier on good fundamentals, not a fix for poor sizing data or misleading product pages. Measure returns against a control group rather than assuming improvement.
For more information, see GOV.UK.