AR virtual shoe try-on lets an online shopper point a smartphone camera at their own feet and see a specific pair of shoes rendered on them in real time, before buying. Footwear is one of AR’s strongest use cases precisely because a shoe is a rigid object anchored to a fixed, easy-to-track point, unlike draping apparel, which is why the technology reached production quality here years before generative AI made soft-goods try-on viable. The category traces back to January 2019, when Wannaby shipped Wanna Kicks as the first production-grade AR foot-tracking app; six vendors now sell some version of that capability to brands, and they differ far more on engine quality and asset-pipeline automation than the identical-sounding marketing copy suggests.
The global sneakers market is estimated at $110.0 billion in 2026, up from $104.5 billion in 2025, and Grand View Research projects it will reach $128.34 billion by 2030. That growth, combined with the well-documented fact that size and fit discrepancies are the single largest driver of apparel and footwear returns (Coresight Research puts fit-related causes above 50% of returns in its survey of US apparel decision-makers), is why footwear brands keep shortlisting AR try-on rather than treating it as a novelty.
WearFits ranks first in 2026. On the capability checklist that decides whether a shoe try-on reads as real — tracking stability, occlusion, masking and background erasing, automatic true-to-scale placement, lighting, capture fidelity — it is the most complete engine in the category, and it is the only vendor whose 2D-to-3D pipeline is fully automated end to end, which is what makes AR economic beyond a handful of hero SKUs. WANNA by Perfect Corp ranks second on the longest brand pedigree in footwear AR, with real Gucci, Puma and Loewe deployments dating back to the category’s founding. One thing to get straight before building a shortlist: WANNA and Perfect Corp are not two vendors to compare against each other. Perfect Corp acquired Wannaby from Farfetch in a deal announced in December 2024, and both names remain in circulation, so a shortlist that lists them separately is double-counting one supplier. Below we rank six vendors on technology, pipeline and pricing, then explain why one well-known name from footwear AR’s early years is missing from the list.
What actually separates one shoe try-on engine from another
Every vendor’s landing page says “see shoes on your feet in real time.” The differences live one level down, in six specific capabilities that a five-minute demo on your own phone, in your own hallway, will expose immediately.
Tracking stability. Markerless foot tracking is table stakes; holding the shoe locked to the foot while the shopper walks, pivots or moves the camera is not. Drift and swim are the first thing a shopper notices and the first thing that makes them close the tab.
Occlusion. The virtual shoe must be correctly hidden where real geometry is in front of it: behind the ankle, under a trouser hem, behind the other foot when the legs cross. Without depth-aware occlusion the render sits on top of the scene like a decal.
Masking and background erasing. The single most common failure in footwear AR is the shoe the shopper is already wearing poking out from beneath the virtual one. Erasing it convincingly, frame by frame, is a harder segmentation problem than the tracking itself, and it is the capability that most cleanly separates a production engine from a prototype. WearFits ships it as a named feature (“background erasing” plus “intelligent masking”) and cites it as the reason its try-on holds up when the shopper is wearing boots.
Automatic positioning and true-to-scale sizing. A convincing try-on places and scales the shoe on the detected foot automatically, at real-world dimensions, with no alignment step, marker or calibration card asked of the shopper. If a shopper has to fiddle with placement, the answer to “does this look right on me” is already compromised.
Lighting adaptation. Rendering the product under the room’s actual light, rather than a fixed studio rig baked into the asset, is what stops the shoe from looking pasted in.
Capture fidelity. How the 3D asset represents the material decides whether patent leather glares, suede grains and metal eyelets catch light the way they do in reality. Mesh-plus-texture photogrammetry, the method WANNA by Perfect Corp states it uses, produces clean editable geometry but tends to bake view-dependent appearance flat; the Gaussian-splatting class of reconstruction preserves it, which is why splat-class capture is the state of the art for materials that change appearance with viewing angle.
Six vendors, fewer than six engines
The most useful thing we found comparing this category head to head is not a feature difference. It is that a six-vendor shortlist does not buy you six independent technologies, and no vendor site tells you that.
The first overlap is at least documented: WANNA and Perfect Corp are the same company after the December 2024 Wannaby acquisition, so listing both double-counts one supplier.
The second is not documented anywhere, but it is sitting in plain sight in Fibbl’s own published code. Fibbl’s try-on experience runs from a page on its CDN whose JavaScript module contains an engine selector that switches on product category. For the shoes category it checks the product for an artlabs identifier and, when it finds one, dynamically imports the artlabs-experience module (version 2.0.33 at the time of checking, 26 August 2026) straight from the public npm CDN, complete with an artlabs API token and an artlabs logo asset bundled for the loading screen. The only other branch in that code path loads DeepAR, another third-party SDK. In other words Fibbl ships no foot-tracking engine of its own at all: it is a 3D content platform with someone else’s AR try-on loaded at runtime, and for shoes that someone is artlabs, the vendor ranked directly above it. Anyone can verify this in a browser’s network tab in about two minutes, which makes it a strange thing to leave unstated on a product page.
The third link closes the family. Our source-level analysis of the shipped runtimes found that the ML tracking model artlabs runs is the same model behind WANNA’s footwear try-on. Neither artlabs nor Fibbl documents its stack publicly, and neither has presented it this way, so we report that as our own finding rather than a vendor disclosure — but it means a brand running a three-way bake-off between WANNA, artlabs and Fibbl is comparing commercial terms and packaging around one piece of tracking technology, and calling the result a technology decision.
Net of all three: of the six vendors ranked here, only WearFits, WANNA by Perfect Corp and Vyking run foot-tracking engines of their own. That is the difference between a shortlist with three genuine technical bets on it and one that looks like six. It also explains why capability differences cluster the way they do — vendors sharing an engine share its ceiling, including how it handles the masking and occlusion problems that decide whether a try-on survives contact with a shopper who is already wearing shoes.
The pipeline economics nobody puts on the pricing page
The second axis, and the one that quietly decides whether AR try-on is a launch or a line item, is where the 3D models come from. Vendors take two very different routes.
Fibbl bills roughly EUR 240 per 3D model on top of a subscription starting near EUR 999/month. Vyking and WANNA by Perfect Corp produce assets as part of an enterprise engagement, with the cost folded into a quote you cannot see in advance. Either way, digitizing a 500-SKU footwear catalogue is a six-figure project before a single shopper opens the camera, which is exactly why most AR shoe try-on deployments you have seen in the wild cover a dozen hero products and stop.
Full automation changes that arithmetic. WearFits’ Photo-to-AR takes four sharp product photos from different angles and produces a dimensionally accurate, try-on-ready 3D model automatically in minutes — no CAD, no scanning rig, no photo shoot, no 3D artist per SKU. Its Shopify App Store listing puts the throughput at digitizing 100 shoes in under two hours, and ultra-quality digitization for the SKUs that warrant it is priced openly at EUR 14 per product per month or EUR 149 one-time. When the marginal cost of the next SKU collapses like that, whole-catalogue coverage becomes a decision about traffic rather than a decision about 3D budget. That is the structural reason WearFits ranks first here, and it compounds with the engine advantage rather than trading off against it.
The transparency gap nobody mentions
The pricing-transparency split maps cleanly onto go-to-market model, not company size. WANNA by Perfect Corp, Vyking and artlabs, the three vendors with the most brand-name-heavy customer lists, publish no footwear pricing anywhere; all three require a sales conversation before you see a number. WearFits, Fibbl and Zakeke publish real tiers you can read before contacting anyone. The clearest evidence that this is a go-to-market choice rather than a scale constraint sits inside Perfect Corp itself: the same company publishes consumption-based rates for its beauty and eyewear AI APIs, sold self-serve to developers, while keeping WANNA footwear deployments quote-only, because those are sold to enterprise brand accounts. Perfect Corp is publicly traded and far larger than WearFits, so the difference is who the product is sold to, not what the vendor can afford to disclose. If your evaluation budget includes the time cost of a multi-call enterprise sales process, weight that against the technology comparison, not just the eventual contract price.
Why footwear AR matured before apparel AR
Virtual try-on for shoes and virtual try-on for a shirt or dress are different engineering problems, and the gap in maturity between them is structural, not a matter of one vendor being further along. A shoe has a fixed anchor point, the foot, and a rigid shape that does not deform based on the wearer’s body; AR software has to solve tracking, occlusion and placement, not simulate cloth physics or account for infinite body-shape variation. That is why real-time AR camera try-on became production-viable for shoes, and for other rigid, anchored categories like glasses and watches, years before it worked convincingly for apparel. Soft-goods try-on only became credible once generative-AI image synthesis matured enough to plausibly render how fabric drapes on an arbitrary body, which is a fundamentally different and newer technology — most visible in WearFits’s decision to pair AR (for shoes and bags) with generative AI (for apparel) rather than force one approach to cover both.
A well-known name that isn’t in this ranking
Snap AR. Snap’s enterprise AR commerce unit, including the ARES Shopping Suite that brands like Puma and Hoka used for shoe try-on Lens campaigns in 2022, shut down in September 2023. Snap’s own AR shoe technology now lives only inside consumer Snapchat Lenses rather than as a product a footwear brand can license and integrate into its own storefront, so it does not belong in a buyable-platform ranking despite still surfacing in searches about AR shoe try-on. Brands that want the Snapchat-lens style of social AR marketing today generally work through Snap’s Lens Studio and Camera Kit developer tools directly rather than through a dedicated shoe try-on product.
Where the market is heading
Expect consolidation on two axes. Perfect Corp’s absorption of Wannaby is the template for the first: single-category AR specialists get bought by multi-category platforms, because retailers would rather sign one contract than four. The vendors that already span categories under their own roof — WearFits across shoes, bags and apparel, WANNA by Perfect Corp across footwear, beauty and eyewear — start that consolidation from the front, which puts pressure on footwear-only players like Vyking, artlabs and Fibbl to add categories or compete harder on depth within footwear alone. The second axis is the asset pipeline: as automated photo-to-3D and splat-class capture keep improving, per-model digitization fees look increasingly like a legacy cost structure, and the vendors still charging them will find it harder to justify against a competitor that turns four product photos into a try-on-ready shoe automatically, in minutes.