The single most useful thing I can tell a fashion brand about AI virtual try-on software is that it is two different markets wearing the same name. One market is tools you can license today: AIUTA, True Fit, Sizebay, Virtusize, Fit Analytics, Bold Metrics, FASHN.ai, Tryva and Zakeke all sell to brands. The other is a feature that runs inside somebody else’s app, where you are a participant rather than a customer. Google is the biggest player in the second category, and it auto-enrols you.
That distinction changes what you can promise your customers, what you can measure and who controls your roadmap. Most roundups I’ve read this year blur it, and several list products that no longer exist.
The short answer: license it, or you are renting it
A brand can license commercial virtual try-on software today, but it cannot license Google Shopping’s Try-on, and it cannot change how Google’s version behaves. Google’s merchant documentation is unusually blunt about the second point: the generated image “can help people find how an item might look, but it doesn’t indicate fit, suggest a size, or indicate size availability from the merchant.”
Everything else follows from that sentence. If you want a renderer, you buy one. If you want a size recommendation, buy a sizing engine, which is a different product with different evidence behind it. If you just want your garments to render on Google for free, you have already got it, whether you asked for it or not.
The strongest-sounding numbers in this category come almost entirely from vendors. No retailer-published measurement cited here isolates a return reduction caused by virtual try-on alone.
The evidence behind the comparison
Product descriptions and prices are from vendor pages checked on 16 October 2026. Garment fidelity, drape and latency were not tested hands-on; performance claims are attributed to vendors or research papers.
The NRF returns figures cited here were published on 15 October 2025. They provide market context, not a measure of what a try-on product will save your brand.
The business case: what returns actually cost
Returns are the reason anyone buys this, and the latest published US figure is $849.9 billion for 2025. The National Retail Federation’s 2025 Retail Returns Landscape, run with Happy Returns and dated 15 October 2025, projected total returns of $849.9 billion, equal to 15.8% of annual sales.
The same report and its accompanying press release give the numbers a fashion brand actually cares about:
- 19.3% of online sales were expected to be returned in 2025, against 15.8% across all channels.
- 17% of holiday sales were expected back, consistent with prior years.
- 9% of all returns were fraudulent, according to retailers surveyed.
- 82% of consumers called free returns an important purchase factor, up from 76% a year earlier.
- 85% of retailers said they were using AI to detect or prevent return fraud.
Returns are not a side effect of e-commerce. At 19.3% of online sales, every point of fit-related return reduction is a direct margin line, and almost none of the published evidence for try-on names its denominator.
Two cautions. These are retailer estimates, not audited transaction data, and they cover all categories. There is no clean apparel-only or online-apparel-only carve-out in the report with a stated method. Any vendor or article that gives you a “fashion returns are X%” number derived from this without explaining the method is guessing.
What a brand can license today: the AI virtual try-on software shortlist
These nine commercial products offer different functions. Note the middle column: several products named “virtual try-on” in roundups do not render anything at all.
| Product | What it actually does | Public pricing (16 Oct 2026) | Input it needs |
|---|---|---|---|
| AIUTA | Single-item and full-outfit try-on; licensed by ASOS for its Feb 2026 launch | Sales only | Product images + customer image or AI likeness |
| True Fit | Sizing agent, now with a “V-TON” visual preview alongside its size guidance | Sales only | Catalog, purchase and return data |
| Sizebay | Sizing core; visual try-on sold as an “exclusive add-on” | Sales only | Height, weight, age + garment measurements |
| Virtusize | Sizing and fit comparison against clothes a shopper owns | $1 first month, then from $39/mo | CSV, Excel or PDF size data |
| Fit Analytics | Fit Finder sizing + AI Shopping Assistant | Sales only | Size and fit preference questions |
| Bold Metrics | Virtual Sizer API using a “digital twin” of 50+ body measurements | Sales only | Self-reported inputs, no photo |
| FASHN.ai | Generative try-on API and studio; Try-On Max and v1.6 endpoints | $19 / $49 / $99 per month | Person image + garment image |
| Tryva | Try-on widget plus AI model generation; one-line install | $29 / $69 / $139 per month | Flat lay or hanger shot |
| Zakeke | 3D configurator with Virtual Try-On as a paid add-on | $59/mo from, VTO from about $40 extra | 3D models, QR or iframe embed |
Two of these deserve a note because their positioning has moved.
Zaker’s try-on is not an apparel body-drape tool. Its own Virtual Try-On page describes “glasses, jewelry, hats, and more,” with AI handling head size and pupillary distance. That is a legitimate product, but it is not what an apparel brand needs.
Vue.ai is no longer primarily a try-on vendor. Its homepage leads with “Enterprise AI Orchestration Platform” and lists virtual dressing rooms as one use case among dozens. If a competitor roundup still describes Vue.ai as a fashion try-on specialist, that page is stale.
What you cannot license, and already have
Google’s Try-on is free, automatic, and not yours. From Google’s own Merchant Center help page:
- “Any merchant with a shopping feed and who is eligible to show free listings is automatically opted into this feature.”
- Supported categories are shoes, tops, bottoms and dresses. Lingerie, bathing suits and accessories are excluded.
- It runs in 18 listed markets including the US, UK, Canada and Australia.
- It requires the user to be logged in and 18 or over, with activity and personalisation enabled.
- It is not available on Shopping Ads, so sponsored results get no Try-on button.
There are two separate Google features and they behave differently. The model-based one, documented in About apparel virtual try-on, uses models that are “real humans, shot professionally by Google” in sizes XXS to XXXL, and does appear in Shopping ads. The shopper-photo version, in Google Shopping Help, lets a user upload their own image or, for US shoppers, generate a digital self from a selfie.
The opt-out is partial, and the documentation is explicit: it “only removes the ‘Try-on’ button from your listings but doesn’t affect listings from other merchants selling the same product.” A competitor stocking your SKU can carry a try-on of your garment whether or not you consented.
There is also a precedent worth sitting with. Google shipped a standalone try-on app called Doppl in June 2025 and shut it on 30 April 2026, announcing the move on X in March 2026 and folding the technology into Search and Shopping. The labs.google/doppl page now returns a 404. The capability survived; the product did not. That is what depending on someone else’s roadmap looks like in practice.
What Google will not do is tell a shopper what size to buy. Its documentation says the render “doesn’t indicate fit, suggest a size, or indicate size availability,” and advises users to “always refer to standard size charts, customer reviews, and product details on the product page.” Any tool whose business case rests on “improve fit accuracy” is therefore not competing with Google. It is doing something Google explicitly declines to do.
The image specification is the real integration job
Google publishes a hard specification for the images it renders from, and your catalogue either matches it or renders badly. Check these requirements before uploading product photos.
From Google Merchant Center Help, the try-on image requirements and best practices:
- At least 512 × 512 pixels, ideally 1024 or higher.
- The image should contain only the listed garment.
- It should show the entire garment as a full-body picture.
- The garment is on a model facing directly forward with arms down to the side.
- Lay-flat garments should avoid excessive folds or wrinkles.
- Garment details should not be covered by hands, handbags or accessories.
- For tops: sleeves rolled down, hoods down. The model-based variant adds zippers closed and tops untucked.
Read that as a shot list. One front-facing, full-body, arms-down image per SKU. If your catalogue is flat lays and three-quarter editorial crops, your listings will render worse than a competitor whose photography matches the spec: on an experience neither of you controls.
That is the real argument for spending money on source imagery rather than on the widget. The on-model photograph is the input to Google’s renderer, to any vendor tool you add, and to your own product page. It is the only asset in that chain you own outright.
2D diffusion versus 3D simulation
Choose 2D diffusion for apparel catalogues, and 3D simulation only where the garment justifies the asset cost. The two architectures fail differently.
2D diffusion synthesises a picture from flat artwork plus a body image. It needs only the product shots you already have, returns in seconds, and scales to hundreds of SKUs cheaply. Tryva advertises 12–18 second results and roughly $0.29 per image on its entry plan. What it cannot do is tell you how a bias-cut dress falls or whether a size two sizes smaller will pull at the hip, because nothing in the pipeline simulates fabric.
3D simulation requires a modelled garment mesh plus body data and renders drape physically. Shopify’s enterprise guide on virtual fitting rooms, published December 2025, gives useful operational ceilings: keep each AR model under 5MB, keep 3D apparel models under 50,000 polygons, and target scene loads under two seconds on mobile.
That is why 3D wins in accessories and loses in apparel. Vyking describes launching virtual try-on for 3,000 models across 30-plus markets through its app integration, and a Dr. Martens web deployment. Fittingbox has digitised 195,000-plus frame references across 1,200-plus brands. A pair of sunglasses is one asset. A dress collection is five hundred.
The open-source trap
The two most-cited open virtual try-on models are released under CC BY-NC-SA 4.0, which permits non-commercial use only.
The CatVTON repository states: “All the materials, including code, checkpoints, and demo, are made available under the Creative Commons BY-NC-SA 4.0 license… for non-commercial purposes.” It is a genuinely good model: 899.06 million parameters, 49.57 million trainable, under 8GB of VRAM at 1024 × 768, accepted at ICLR 2025, and it is not licensed for your storefront.
The IDM-VTON repository carries the same licence and is the basis of the ECCV 2024 paper on authentic virtual try-on in the wild.
No open alternative with confirmed commercial-use rights is included here. If you want to self-host, read the licence file before you build. This is a five-minute check that can save a legal problem later.
What it costs, and how vendors charge
Published prices cluster between $19 and $139 per month for self-serve tiers, and go opaque above that. These are the vendor rates listed on 16 October 2026.
| Vendor | Entry | Mid | Top | Public or sales-only |
|---|---|---|---|---|
| FASHN.ai | $19/mo, 200 credits | $49/mo, 750 + 50 daily | $99/mo, 1,500 + 100 daily | Public; API tier sales-led |
| Tryva | $29/mo, 100 credits | $69/mo, 300 credits | $139/mo, 800 credits, API, white-label | Public |
| Virtusize | $39/mo after a $1 first month | Priced to assisted conversions | Custom above 10,000 monthly orders | Public with enterprise gate |
| Zakeke | $59/mo annual, 10 products | $110/mo, 50 products | $255/mo, 100 products | Public; VTO add-on from ~$40 |
| Sizebay | Quote | Quote | Quote | Sales-only, traffic-tiered |
| Fit Analytics | Quote | Quote | Quote | Sales-only |
| Bold Metrics | Quote | Quote | Quote | Sales-only |
| True Fit | Quote | Quote | Quote | Sales-only |
| AIUTA | Quote | Quote | Quote | Sales-only |
Two things this table exposes.
Try-on is frequently an add-on, not a bundle. Zakeke’s own pricing comparison lists Virtual Try-On as “starting from $40” on top of Starter, Grow and Scale. A competitor blog claimed try-on and AR viewing are included across Zakeke’s plans. Zakeke’s pricing page contradicts that directly, and Zakeke’s page wins.
Virtusize’s homepage and FAQ disagree with each other. The hero panel claims a 20.1% average return-rate reduction and 5.1× conversion uplift across 500-plus brands. The FAQ on the same page says “5× conversion uplift and 14% return rate reduction.” The page does not reconcile those figures. Bold Metrics has a similar problem: one panel on the Virtual Sizer page says a 17% average reduction in fit-related returns, and a checklist further down the same page says 18%.
For FASHN.ai, note that the $19/$49/$99 tiers are the studio product sold in credits. Its API documentation recommends Try-On Max for most applications and offers Try-On v1.6 for “real-time, cost-sensitive integrations,” but API pricing is not published. Any article quoting “$19/month for enterprise virtual try-on API” has read the wrong page.
How to roll this out across a 500-SKU catalogue
Roll out in the order of dependency, not in the order of excitement. Image specification first, because everything downstream consumes it.
-
Audit your on-model photography against Google’s published spec. Full body, front-facing, arms down, one garment, at least 512 × 512 and ideally 1024+, nothing occluding the garment. Fix the catalogue before you talk to any vendor, because the same asset feeds Google’s renderer, a vendor widget and your own product page.
-
Pick one high-visibility category, not the whole catalogue. Tops or eyewear. Measure the return rate on that category first, because a tool that fails to move a number you already know is easier to abandon.
-
Instrument before you switch anything on. Track add-to-cart, checkout conversion and return rate as three separate events. Returns take months to read, so the first two give you signal in weeks and the third needs the plumbing in place on day one.
-
Run a matched control. Pick comparable SKUs with try-on disabled and compare like with like. A return rate that falls after launch is not evidence, because returns move with season, mix, promotion depth and returns-policy changes. A vendor with no holdout has not measured anything.
-
Read the data terms before you upload unreleased designs. Some vendors state they never train on customer images; others are vague. Tryva says its guarantee is contractual. Google’s Try-on says it “doesn’t collect or store any biometric data” and does not use photos for training. Get the equivalent in writing from anyone else, and check whether they retain rights to your garment imagery.
-
Ask for the denominator, in writing. “40% fewer returns” means nothing until you know whether it is size-related or total, of enabled SKUs or the whole catalogue, of try-on users or all users. True Fit’s homepage states up to a 40% reduction in fit-related returns when guidance is followed, which is more carefully scoped than most, and it also claims a 1–2% sitewide conversion lift measured against a control group. That second claim is the only properly designed number I saw on any vendor page.
-
Budget the data work, not just the licence. Shopify’s rollout guide suggests standardising size-chart attribute names across SKUs, capturing front, side and back views in neutral lighting, and mapping colour codes and materials to your chosen app’s required fields. That work is unglamorous and it decides whether the tool works.
What breaks: failure modes and gotchas
The most common failure is not a rendering artefact. It is a measurement failure. Try-on users are, by definition, higher-intent shoppers. Compare try-on users to non-users and you are measuring self-selection, not the feature. This inflates almost every published number in this category.
Other things that break:
- Vendor churn is real and fast. The domain drape.ai was listed for sale at $60,000, yet DrapeAI appeared in multiple 2026 roundups as a live recommendation. Reblium, once listed as an AR try-on vendor, is now a browser-based character and hair studio in closed beta. zyler.com now serves Bridely, a bridal-gown try-on product. Vue.ai now leads with enterprise AI orchestration. Verify the vendor still sells what the article says it sells.
- Category coverage is narrower than the marketing. Google excludes lingerie, bathing suits and accessories. Zakeke’s try-on is glasses, jewellery and hats. Always check your actual category against the tool’s supported list before counting it as a fit.
- Sizing engines do not render, and renderers do not size. Sizebay is explicit that its visual try-on “doesn’t calculate which size to buy,” and sells the two as different products. Bold Metrics advertises accuracy “with no scanning, photos, or measuring tapes.” Know which one you bought.
- Onboarding time varies enormously. Sizebay states an average of 30 days from contract to go-live, handling about 95% of the work itself. Virtusize claims two script tags and under three minutes. Both are vendor claims, but the gap between them is the gap between a sizing engine and a 3D pipeline.
- The dramatic retention numbers deserve scepticism. DRESSX, reporting on what it calls its Intelligence Report across 1.2 million luxury shoppers in 216 countries and 83 languages, claims 44% Day-30 retention among try-on users against 1% among non-users. A 44-to-1 gap is more likely to indicate a measurement or cohort artefact than a product effect. No independent audit is cited here.
Legal: consent, body data and generated imagery
This is the part most vendor pages skip, and in 2026 it got sharper, not softer. Treat the following as a starting point for a conversation with counsel, not as advice.
Model and shopper likeness consent. If your catalogue is shot with real models, your release forms need to cover digital manipulation, not just photography. Google states plainly that users “may only upload an image of yourself or an image of someone else that you have permission to use,” and that the user “is solely responsible for ensuring they have the full consent and permission of any individual whose photo you upload.”
GDPR and body measurements. Body measurements are not on the Article 9 special-category list. Article 4 defines biometric data as data resulting from specific technical processing of physical or physiological characteristics that allows or confirms unique identification, and defines data concerning health as data revealing health status. A waist measurement is neither. Two things move it: building a face or body template intended to identify a person, and drawing inferences about health or body composition. The EDPB’s SME guide confirms that photos and video of individuals are personal data and lists the special categories. Get advice on the specific product’s processing, not on the category label.
EU AI Act Article 50. Article 50 has applied since 2 August 2026, per the European Commission’s regulatory framework page, which also records that the AI Act became applicable on that date and that the AI Omnibus entered into force on 27 July 2026. The article text requires providers of systems generating synthetic image content to mark outputs as artificially generated in a machine-readable format, and requires deployers to disclose content constituting a “deep fake”: defined in Article 3(60) as AI-generated or manipulated content resembling existing persons, objects, places, entities or events that would falsely appear authentic. Where content forms part of an evidently artistic, creative, satirical, fictional or analogous work, the obligation narrows to disclosing that such content exists. A realistic invented model, or a real product shown more flattering than in life, is exactly where this gets contested. No enforcement decision on product imagery is cited here.
US advertising claims. The FTC’s advertising guidance is short and unforgiving: “claims in advertisements must be truthful, cannot be deceptive or unfair, and must be evidence-based.” If you write “find your perfect size,” you need evidence. If you write “see yourself in this dress,” you are describing a visualisation, and you should describe it that way on the page.
Accessibility. The European Accessibility Act explicitly covers e-commerce among the products and services in scope. A camera-permission-gated try-on button is unusable for a meaningful number of shoppers, so ship a static fallback: your own on-model imagery, an accurate per-garment size chart and the fit feedback you already collect.
The bottom line
Spend first on the image, second on the size recommendation, and last on the widget. That is the ordering the available evidence actually supports, even though it is the reverse of how vendors pitch.
In 2026 the renderer is close to free and largely out of your hands. Google gives it to every eligible merchant automatically and will keep improving it whether or not you are in the market. What is still scarce, still expensive and still yours to choose is a trustworthy size answer, a per-garment size chart and an on-model photograph that matches a published specification.
If you buy anything, buy it for the job Google refuses to do. Google says its render “doesn’t indicate fit, suggest a size, or indicate size availability from the merchant.” That sentence is the opening for every sizing engine on the market, and it is also the sentence you should quote when a vendor promises you a 40% return reduction.
And when a vendor gives you a number, ask for the denominator, the holdout and the customer base. Three questions. If they cannot answer all three in a sentence each, you are looking at a case study, not a measurement.
FAQ
What is AI virtual try-on software?
It is software that generates an image of a person wearing a specific garment without photographing them in it. Most current products fall into two families. Diffusion-based tools combine a shopper’s photo or your flat-lay product shot with a generative model. 3D garment simulation tools require a modelled garment mesh and avatar, then physically simulate drape. A third family, size recommendation engines, predicts the right size rather than drawing a picture at all. Google’s own documentation says its try-on “doesn’t indicate fit, suggest a size, or indicate size availability from the merchant,” which is the cleanest public statement of what the visual layer does and does not do.
Does virtual try-on actually reduce returns?
There is directional evidence that fit tooling reduces size-related returns, but no retailer-published figure cited here attributes a large reduction to virtual try-on alone. The NRF’s most recent published figures put returns at $849.9 billion in 2025, 15.8% of annual sales and 19.3% of online sales, published October 2025. The widely repeated “25 to 40% fewer returns” figure appears on vendor pages without a traceable retailer source. Zalando’s 8% size-related figure and ASOS’s 160 basis point improvement are both reported by third parties; neither is confirmed against the companies’ own publications.
How much does virtual try-on software cost?
Public entry points I verified on 16 October 2026: FASHN.ai Basic at $19/month for 200 credits, Pro at $49/month for 750 credits plus 50 daily, Agency at $99/month for 1,500 credits plus 100 daily, with top-ups at $0.10 per credit. Tryva charges $29/month for 100 credits, $69 for 300 and $139 for 800. Virtusize is $1 for the first month then from $39/month, tied to assisted conversions. Zakeke starts at $59/month but sells Virtual Try-On as an add-on from about $40. Sizebay, Fit Analytics, True Fit, Bold Metrics and AIUTA do not publish pricing.
Can I use Google virtual try-on on my own store?
No, not as an embeddable widget. Google’s feature renders inside Google Search, Google Shopping and Google Images, and any merchant with a shopping feed eligible for free listings is automatically opted in. You cannot place it on your own product page, and opting out requires contacting Google Customer Support from Merchant Center. Google’s help page also notes that opting out “only removes the ‘Try-on’ button from your listings but doesn’t affect listings from other merchants selling the same product.” The shopper-facing version is not available on Shopping Ads.
What is the difference between 2D diffusion and 3D garment simulation?
2D diffusion tools take flat artwork and a body image and synthesise a picture. They need only the product shots you already have, which is why they are cheap and fast, and they cannot answer a question about drape under gravity or how a size will pull. 3D simulation requires a modelled garment mesh plus body data, then renders fabric behaviour physically. Shopify’s own guide suggests keeping 3D apparel models under 50,000 polygons and each AR asset under 5MB. The trade-off is asset cost: 3D is defensible for a footwear or eyewear range and painful across a 500-SKU apparel catalogue.
Can I self-host an open source virtual try-on model commercially?
Not with the two most-cited models. The CatVTON repository states its code, checkpoints and demo are released under CC BY-NC-SA 4.0, which permits non-commercial use only, and IDM-VTON’s repository carries the same licence. CatVTON is genuinely lightweight, at 899.06 million parameters and under 8GB of VRAM at 1024 by 768 resolution, but running either inside a commercial storefront breaches the non-commercial term. Read the licence file before you build, not after a revenue review.
Is body measurement data special category under GDPR?
Not by itself. Article 4 of the GDPR defines biometric data as data resulting from specific technical processing of physical or physiological characteristics that allows or confirms unique identification, and defines data concerning health as data related to physical or mental health that reveals health status. Article 9 lists health data and biometric data for unique identification as special categories, and a waist measurement is on neither list. Two things change the answer: a face or body template built to identify a person, and any inference the vendor draws about health or body composition. Get counsel on the specific product, not on the category.
Do AI-generated on-model product images need a disclosure label in the EU?
Possibly. Article 50 of the AI Act has applied since 2 August 2026. It requires providers of systems generating synthetic image content to mark outputs as artificially generated in a machine-readable format, and requires deployers to disclose content that constitutes a deep fake, which Article 3(60) defines as content resembling existing persons, objects, places, entities or events that would falsely appear authentic. There is a narrower obligation for evidently artistic or fictional work, limited to disclosing that the content exists. No enforcement decision on product imagery is cited here.
What is the best virtual try-on for Shopify?
It depends on which of two jobs you need. Shopify’s own enterprise guide lists Fitnonce for AI sizing from a customer photo, Swan AI Fitting Room for avatar-based try-on, Camweara for AR accessories and YouCam Virtual Try-On for beauty. Shopify separately says pricing across the category “range[s] from $129 to $549/month.” If you want shopper-facing apparel try-on, Shopify’s list has no equivalent of Tryva or FASHN.ai, both of which install by one script tag. Shopify also published a 30-day rollout plan worth stealing regardless of vendor.
How do I roll out try-on across a 500-SKU catalogue?
Start with one category, not the catalogue. Fix your on-model image to Google’s published specification first, because that same image is the input to Google’s try-on, to any vendor tool and to your own product page. Measure against a matched control group of SKUs with try-on disabled, tracking add-to-cart, checkout conversion and return rate separately. Returns take months to read, so instrument the first two immediately and set the third up before launch. Sizebay states an average of 30 days from contract to go-live with the vendor handling about 95% of the work, which is a reasonable benchmark to hold others to.
Sources
- How Google’s Try-on tool works: Google Merchant Center Help: auto opt-in, category limits, image specification, opt-out scope, and Google’s own statement that the render does not indicate fit or suggest a size
- How the Google Try-On tool works: Google Shopping Help: shopper eligibility, photo requirements, Nano Banana digital self, Google’s privacy position and consent responsibility
- About apparel virtual try-on: Google Merchant Center Help: model sizes XXS to XXXL, real Google-shot models, Shopping ads eligibility, tops-specific image rules
- Virtually try on clothes with a new AI shopping feature: Google: June 2023 launch with Anthropologie, Everlane, H&M and LOFT, plus Google’s own representation and satisfaction survey figures
- Generate a digital version of yourself to virtually try on clothes: Google: December 2025 selfie-to-full-body feature using Nano Banana (Gemini 2.5 Flash Image)
- Google Labs on X: Doppl shutdown announcement: primary confirmation that the Doppl app shut down on 30 April 2026 and moved into Search and Shopping
- Google shut down Doppl. Here is what it means for virtual try-on: Corlen: competitor write-up of the timeline; used only for chronology, not for claims about Google’s product
- 2025 Retail Returns Landscape: National Retail Federation: $849.9 billion, 19.3% of online sales, 82% free-returns importance, 9% fraud; published 15 October 2025
- Consumers Expected to Return Nearly $850 Billion in Merchandise in 2025: NRF press release: 15.8% return rate, 17% of holiday sales, 85% of retailers using AI for fraud detection, Gen Z return frequency
- NRF and Happy Returns 2024 Consumer Returns in the Retail Industry: $890 billion and 16.9% for 2024, used as the prior-year comparison; published December 2024
- ASOS launches hybrid approach to virtual try-on: ASOS plc: 17 February 2026, AIUTA partnership, around 10,000 products, iOS app, select UK and US customers, 4–7 second loads
- AIUTA: fashion AI for virtual try-on and product imagery: single-item and full-outfit try-on positioning; used to confirm the vendor is a licensable commercial product, not a research demo
- True Fit: agentic shopping platform with native virtual try-on and MCP fit intelligence; vendor claims of 1–2% conversion lift against a control group and up to 40% fit-related return reduction, plus 72% market share and $616 billion transaction value claims
- Sizebay Virtual Fitting Room: sizing-first product with visual try-on as an add-on, 1,200-plus stores, 30-day average implementation, and Sizebay’s own distinction between its B2B tool and Google’s
- Virtusize and Virtusize pricing: $1 first month then from $39/month, conversion-tied pricing, platform coverage, SOC 2 Type II; hero and FAQ figures conflict
- Fit Analytics Fit Finder and AI Shopping Assistant: sizing claims and platform-agnostic positioning; vendor-reported conversion and return figures
- Bold Metrics Virtual Sizer: digital-twin sizing with no photo or tape input; vendor-reported conversion, AOV and return figures that conflict between panels of the same page
- FASHN.ai pricing and FASHN API documentation: published credit tiers and the split between Try-On Max and Try-On v1.6; API pricing is not published
- Tryva: $29/$69/$139 tiers, credit volumes, 12–18 second results, one-line install, and the contractual no-training claim
- Zakeke pricing and Zakeke Virtual Try-On: plan tiers, product limits, transaction fees, Virtual Try-On as a paid add-on, and the accessory-focused scope of the tool
- Botika: AI fashion model generation at scale; vendor-reported production, turnaround and conversion figures
- Stylitics: current positioning as a retail AI platform for outfitting, imagery and catalog enrichment, with no shopper-facing try-on listed
- Vyking: 3D platform, Dr. Martens deployment, 3,000 models across 30-plus markets
- Fittingbox: eyewear specialist, 195,000-plus digitised references, 1,200-plus brands, patent history
- Style.me: 3D virtual fitting room and AR try-on for footwear and accessories
- Reblium: now a browser-based character and hair studio in closed beta; used as evidence of vendor churn in this category
- ModiFace: a L’Oréal company, beauty-category virtual try-on, useful only as a contrast case for apparel
- drape.ai: the domain was displaying a Spaceship “domain for sale” listing at $60,000 on 16 October 2026, while the brand was still recommended in 2026 roundups
- zyler.com: the domain served Bridely’s bridal try-on product on 16 October 2026 rather than a general fashion virtual try-on product
- CatVTON repository and paper: ICLR 2025 model, 899.06M parameters, sub-8GB VRAM, and the CC BY-NC-SA 4.0 non-commercial licence
- IDM-VTON repository and paper: ECCV 2024 model and the same CC BY-NC-SA 4.0 licence
- Shopify: Virtual Fitting Rooms: A Retailer’s Guide for 2026: the $129–$549 monthly pricing band, 3D asset ceilings, app evaluation criteria, and the four-week rollout plan
- GDPR Article 4: Definitions and Article 9: the biometric data and health data definitions and the special-category list
- EDPB SME Data Protection Guide: Data protection basics: confirmation that photographs of individuals are personal data and how special categories are treated
- European Commission: AI Act regulatory framework: applicability from 2 August 2026, transparency rules taking effect in August 2026, AI Omnibus in force 27 July 2026
- Article 50: Transparency Obligations (EU AI Act text explorer): machine-readable marking, deep fake disclosure duty, the Article 3(60) definition and the artistic-work limitation; third-party tracker used alongside the Commission’s page
- European Accessibility Act: European Commission: e-commerce named among covered products and services
- FTC: Advertising and Marketing: the requirement that advertising claims be truthful, non-deceptive and evidence-based
- FTC: Endorsements, Influencers, and Reviews: disclosure obligations relevant to influencer-led try-on campaigns
- DRESSX: 25 AI Virtual Try-On Statistics: the DRESSX Intelligence Report figures; vendor-reported, and the Day-30 retention claim specifically is treated here as unreliable
- Does Virtual Try-On Reduce Returns? The 2026 Evidence: DesignerBox: used as a lead to the Zalando, ASOS and CNBC claims; those three could not be confirmed on primary sources on 16 October 2026 and are flagged as unverified in this article
- Top 10 Virtual Try On Software for Fashion Retailers Compared (2026): QArt Solutions: read as a competitive-landscape sample; vendor-authored and used only to identify names to verify
- Best Virtual Try-On Software for Fashion Brands in 2026: tryva.ai: read for the Zeekit status claim and the Zakeke bundling claim; the latter is contradicted by Zakeke’s own pricing page
