Five Virtual Try-On Launches In Six Months Is A Pattern

December 2025: 𝐙𝐚𝐫𝐚 rolls out VTO in Mexico, the UK, Germany, the Netherlands, and Italy. By spring 2026 the company reported more than seven million try-on sessions across 43 markets.

May 7, 2026: OTB Group launches virtual try-on on 𝐃𝐢𝐞𝐬𝐞𝐥 and 𝐉𝐢𝐥 𝐒𝐚𝐧𝐝𝐞𝐫 in the US and Europe, with 𝐌𝐚𝐫𝐧𝐢 and 𝐌𝐚𝐢𝐬𝐨𝐧 𝐌𝐚𝐫𝐠𝐢𝐞𝐥𝐚 queued behind. Built in partnership with Google Cloud.

May 27, 2026: 𝐊𝐦𝐚𝐫𝐭 Australia launches an AI shopping assistant called Joy. Virtual try-on for apparel is the headline feature. First of its kind in Australian retail.

Add to that list 𝐖𝐚𝐥𝐦𝐚𝐫𝐭’𝐬 Be Your Own Model program expanding across 270,000 apparel SKUs, 𝐀𝐦𝐚𝐳𝐨𝐧 shipping virtual try-on for shoes across New Balance, Adidas, Reebok, Puma, and Asics, and 𝐆𝐨𝐨𝐠𝐥𝐞 𝐒𝐡𝐨𝐩𝐩𝐢𝐧𝐠 rolling its consumer-facing VTO into new countries last October.

One launch is a pilot. Two is a coincidence. Five public launches in six months across mass-market, premium fashion, and the retail giants are waking up.

Here is what is happening, why it is happening now, and where 𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 fits.

𝐖𝐡𝐲 𝐧𝐨𝐰

Three things had to be true simultaneously for apparel VTO to cross from pilot to production at scale.

The first was model quality. Diffusion-based generative models capable of rendering a specific garment on a specific body with correct drape, fabric behavior, and lighting only became usable in the last 18 months. Earlier VTO efforts either looked synthetic, required a controlled-studio image of the shopper, or limited the catalog to a few hundred curated items. The current generation handles arbitrary catalog imagery as input and produces a render the shopper actually recognizes as themselves.

The second was the device. 81% of fashion eCommerce traffic and 72% of fashion eCommerce sales are mobile in 2026. Try-on becomes interesting only when the camera, the shopper, and the cart are in the same device. Walmart, Zara, Kmart, Amazon: every 2025 to 2026 launch I’ve tracked is mobile-first.

The third was the shopper. 44% of Gen Z and 46% of Millennials already use AI as part of their shopping process, and 90% and 95% respectively say AI improves the experience. This is a generation that does not need to be educated on why an avatar try-on is worth a tap. They expect it.

When model quality, device adoption, and shopper readiness arrive in the same window, you get five launches in six months.

𝐖𝐡𝐲 𝐦𝐨𝐬𝐭 𝐦𝐞𝐫𝐜𝐡𝐚𝐧𝐭𝐬 𝐰𝐨𝐧’𝐭 𝐛𝐮𝐢𝐥𝐝 𝐭𝐡𝐢𝐬

The brands shipping today are not building VTO. They are partnering with or acquiring tech companies to build it for them.

OTB partnered with Google Cloud. Kmart partnered with Google Cloud. Walmart acquired Zeekit in 2021 and turned it into Be Your Own Model. Amazon built VTO inside AWS and ships it as a Bedrock primitive. Even the buyers with the deepest engineering benches chose an expensive co-dev effort.

The reason is the cost curve. Training a usable apparel-class try-on model requires a labeled triplet dataset of person plus garment plus ground-truth result. Constructing that dataset is the dominant cost and the dominant time sink. Public ML benchmarks put a first-pass exploration at $40K to $50K per month for a small data and ML team, with $10K-plus per month in ongoing infrastructure once a deep-learning model is in production. Time to a production-quality try-on render is two to six months for a fine-tune, longer if the team is building from scratch. That is before the merchant has packaged a UX, integrated a Product Detail Page (PDP) surface, handled mobile camera permissions, and operationalized a feedback loop.

Walmart, sitting on more apparel session data than any retailer outside of Shein, chose to acquire Zeekit rather than build the model in-house. That is the signal. The build-vs-buy answer for any merchant smaller than Walmart was settled in 2021. The 2025-2026 launches just confirm it.

𝐖𝐡𝐚𝐭 𝐭𝐡𝐞 𝐧𝐮𝐦𝐛𝐞𝐫𝐬 𝐬𝐚𝐲

VTO is not a vanity feature. Published apparel data shows 15-20% conversion lifts on try-on sessions, with AR try-on users converting at 2.3x the rate of non-users. Return rate drops up to 40% on apparel orders that went through a try-on session. One published study clocked 24% returns vs. 36% on the non-try-on baseline.

For a Shopify mid-market merchant doing $5M to $50M in apparel GMV with a 25%-plus return rate, those numbers are not features. They are P&L.

𝐖𝐡𝐞𝐫𝐞 𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 𝐟𝐢𝐭𝐬

𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 is our virtual try-on product. It is in MVP today, demo available by appointment. Shopify delivery is on track for July. Shopify will be first because that is where the merchants live who are watching Zara, Diesel, and Kmart and asking “how do I get this in my store by Q4.”

What 𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 is not, is a single-feature point solution. It is one surface of the 𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 platform. The PDP try-on experience matters more when the shopper arrives at the right page in the first place. That’s 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓’𝐬 job. Trimodal AI discovery (natural language, precise color, and image similarity) delivers the shopper to the correct page. Digital-twin try-on lives there. One install, one engine, one merchant relationship.

That combination is the part the standalone VTO vendors cannot match. Veesual, FASHN, Doji, Lalaland: all credible vendors, all serving a narrower slice of the merchant problem. Google Cloud sells the model API and leaves the integration to OTB and Kmart. Amazon sells AWS infrastructure and leaves the deployment to retailers willing to staff for it.

𝐯𝐢𝐬𝐮𝐚𝐥𝐀𝐈 is the only platform shipping AI discovery, catalog hygiene, agentic-commerce distribution to ChatGPT, Perplexity and Google Shopping, and virtual try-on on one engine, for the mid-market merchant who needs to ship in 90 days, not 90 weeks.

𝐀 𝐬𝐡𝐨𝐫𝐭 𝐰𝐨𝐫𝐝 𝐨𝐧 𝐭𝐡𝐞 𝐬𝐡𝐨𝐩𝐩𝐢𝐧𝐠 𝐞𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞

The best-in-class shopping experience in apparel in 2026 is no longer keyword search into a 12-column filter into a static product page. It is conversational, visual, and on a phone. The shopper types “burnt orange wrap dress for a fall wedding.” Trimodal discovery returns the right ten products. The shopper taps the one they like. 𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 renders the dress on their digital twin in five seconds. They add the full outfit to the cart with the confidence that comes from seeing themselves in it.

That is the experience 𝐬𝐡𝐨𝐩𝐩𝐞𝐫𝐆𝐏𝐓 plus 𝐬𝐭𝐮𝐝𝐢𝐨𝐆𝐏𝐓 is built to deliver with a one-click install. The API for accelerating non-Shopify deployments is coming in 2H26.

𝐓𝐡𝐞 𝐭𝐚𝐤𝐞𝐚𝐰𝐚𝐲

If you sell apparel, the question is no longer whether to ship VTO. It is who you trust to ship it with, how to deploy it cost effectively and whether you treat it as a feature bolt-on or as one surface of an AI-native platform you can actually grow into.

Zara and Diesel have made their picks. Kmart has made its pick. The mid-market is up next.