Sunglasses virtual try-on: lens tint, face width, and why eyewear is AR's best case

If you had to pick one product category where augmented reality genuinely settles the buying decision, it would be eyewear. Not furniture, not fashion, not beauty. Sunglasses.

The reason is unglamorous. A frame is a rigid object with fixed dimensions that sits on a part of the body phones track extremely well, and the questions a shopper has about it are geometric rather than chromatic. Everything that makes AR unreliable elsewhere either does not apply or barely matters here. We build eyewear virtual try-on, so take the enthusiasm with the appropriate discount, but the underlying argument holds regardless of who is making it.

The objection try-on removes

Sunglasses online have one dominant objection, and it is well documented. Fittingbox's 2025 report, cited in Shopify's write-up on AR shopping, found that 54 percent of customers who were not ready to buy glasses online wanted to check that the frames fit properly. The same source reports that 29 percent of glasses shoppers have already used virtual try-on at least once.

More than half of the hesitation in the category is one question, and it is a question about fit. That is unusually tidy. In most categories the objections are scattered across price, quality, sizing, delivery and trust. In eyewear there is one big one and it is directly answerable.

Why the technology is up to it here

Three properties line up.

The object does not deform. A frame has a fixed lens width, bridge width and temple length. A 3D model of it is correct everywhere, unlike a garment, which changes shape depending on the body inside it.

The face is the best-tracked surface in consumer computer vision. Landmark detection runs at full frame rate on a mid-range phone and holds through head rotation. Placing a rigid object at the bridge of a nose, at the right scale and angle, is a solved problem.

And the shopper's question is about shape and proportion, not colour. Does this frame suit my face, is it too wide, does the brow line sit right. Those are geometry. The lighting and colour errors that make makeup try-on approximate do not affect the answer.

Face width, which is the real measurement

The part of eyewear try-on that separates a good implementation from a toy is scale.

An overlay that draws a frame on a face without knowing how big that face is can only tell you about style. A correctly scaled try-on estimates the face width from the tracked landmarks and the camera geometry, then renders the frame at its true physical dimensions relative to it. The difference is visible immediately: in a scaled try-on, a 145 millimetre frame looks obviously too wide on a narrow face, which is exactly the information the shopper needs.

This is worth asking any vendor about directly. Do you render at true scale, and how do you estimate face width. Some depth-capable phones measure it; others infer from interpupillary distance. Either can work. No answer at all means you are buying a sticker.

Pair it with the numbers on the page. The three measurements printed on a temple arm, lens width, bridge width and temple length, mean nothing to most shoppers, but shown alongside a scaled try-on they suddenly have context. And if the shopper already owns a pair that fits, letting them type in those three numbers is the cheapest recommendation engine in the category.

Lens tint, the one genuinely hard part

Here is where sunglasses are harder than prescription frames.

A tinted lens is not a coloured surface. It is a partly transparent element that changes what you see through it, and on the wearer's face it does three things at once: it darkens the eyes behind it, it reflects the environment off its front surface, and on a mirrored or gradient lens it does both unevenly across the lens.

Rendering that convincingly means compositing the tint over the tracked eye region with the correct transmission, adding a reflection that responds to head movement, and handling gradients along the lens height. Implementations that skip the reflection produce a flat grey oval that looks nothing like real sunglasses, and shoppers discount the whole try-on because of it.

Mirrored and strongly coloured lenses are the hardest case, because the reflection carries most of the visual character and the environment being reflected is unknown. The practical answer is a plausible studio reflection rather than a real one, plus a product photograph of the lens itself nearby so the shopper can see the true colour. As with any camera-based tool, the colour on screen is an indication, not a match.

What it does to the numbers

We would rather point at mechanism than promise a figure, but the direction is consistent across sources. Shopify's merchant data from September 2020 found interactions with products carrying 3D or AR content converted 94 percent more often than comparable products without it. In Shopify's AR shopping write-up, Rebecca Minkoff reported shoppers 44 percent more likely to add to cart and a 65 percent purchase lift on AR-enabled product pages, and Gunner Kennels cut return rates by 5 percent. None of those are eyewear, so treat them as evidence that the mechanism works rather than as a forecast for your catalogue.

For context on what returns cost, the National Retail Federation and Appriss Retail put United States returns at 743 billion dollars in 2023, about 14.5 percent of retail sales. In eyewear, where more than half of the hesitation is a fit question, the share of that which is addressable is unusually high.

Doing it well, in practice

Put the try-on in the main image gallery, as the first or second slot. Below the buy button it does not get used.

Cover the range, not five hero frames. Shoppers learn very quickly whether a feature is everywhere, and a sparse rollout teaches them not to look.

Model the frames properly. The materials are the job: acetate has depth and internal colour variation, metal has anisotropic highlights, and a frame modelled with flat colours looks cheap next to the product photograph directly beside it.

Keep the file small. A frame should load in a couple of seconds on mobile data or shoppers leave before it appears.

And offer the fallback. Some people will not grant camera access. A rotatable 3D view of the frame in the page, plus the three measurements and a comparison against a pair they already own, serves them without a camera at all.

What we build

We make the 3D models, the web viewer and the eyewear virtual try-on, and we host the assets. The same production line makes configurators and web AR for other categories: Sweef for modular furniture, MELIMELI for a fabric range where a new material is a texture rather than a photo shoot, Contura for placing a product in a real room at true size. The solutions page covers the range and prices are published.

The way to judge any of it is on your own product. Send us one frame and we will build the model and put it in a live viewer, free, and it is yours to keep whatever happens next: free sample.

Sources

• Shopify, AR shopping, including Fittingbox 2025 eyewear figures, Rebecca Minkoff and Gunner Kennels results.
• Shopify, merchant data on 3D and AR content, September 2020.
• National Retail Federation and Appriss Retail, 2023 Consumer Returns in the Retail Industry.