Some retail categories are notoriously tough to buy online unless you know exactly the brand, model, color, and size you need.  When buying products like make-up, eyewear, fitted apparel, shoes, and jewelry, it’s a huge benefit to be able to test out or try-on the physical product prior to purchasing it.  

For example, it’s hard for consumers to know the right shade of make-up without exploring the physical Tester samples at a retail store – a reason Sephora’s try-before-you-buy retail concept has thrived. Consumers are cautious buying expensive jewelry online because they want to see how the chain looks around their neck or how big the diamond ring looks on their hand, leading fine jewelry to have low online conversion rates.  Fitted apparel, like jeans, swimsuits, and tight dresses are risky to buy online without the risk of the return. Eyewear has historically had the lowest eCommerce penetration because it’s so hard to find the perfect fitting glasses for your unique face.

eCommerce adds efficiencies from centralized distribution but often leads to a huge return issue – often 30-50%. Returned items can’t always be resold with only about 76% suitable for resale, 15% going to outlet stores, 7% donated to charity and 1% returned to manufacturers. The reason for all of these returns is driven by poorly fitting items. Poor fit contributes to ⅔ of returns, with returns and reverse logistics costing $100B each year.

Retailers and brands are acutely aware of this pain and have come up with a wide range of solutions that help address these concerns including generous return policies (e.g. Nordstrom’s return policy where they often take things back even after months or years at home) and home try-on programs (e.g. Liingo Eyewear and Warby Parker’s Try 5 on at Home program, Amazon’s Prime Wardrobe program where you aren’t charged until 7 days after you get the items or TryNow, a Shopify plug-in that allows brands to offer try-before-you-buy to customers). There has also been a major push towards more scalable, virtual solutions including the use of augmented reality virtual try-on, commonly referred to as VTO, technology.  

This article goes deeper on virtual try-on to review:

• An overview of virtual try-on
• The key benefits of virtual try-on for customers and retailers/brands
• Key considerations when selecting a virtual try-on
• Virtual try-on leaders by category
• Our predictions on the future of virtual try-on

Virtual Try-On Overview

Virtual try-on technology is a form of augmented reality that superimposes the product onto a customer’s face, body, hand, etc.  This visualization allows customers to see how the products look on them so they can make an informed buying decision.

This is a broad term that can include showing products on a single image, a set of recorded images or a video, or onto a live, real-time camera feed (think Snapchat or Instagram filters). This can even include the creation of an avatar which is used to try-on products.

There are usually few components to the Virtual Try-On:

1. The detection of the face, body part or body which utilizes computer vision to detect various landmarks or pixels on the user
2. Many but not all VTOs reconstruct the face/body part/body in 3D using the detection (generate the shape with proper dimensions); others might just overlay on detected landmark points in 2D image/video/camera feed.
3. The creation of the digital item.  For most VTO solutions today, there is a digital replica of the physical product that is created ahead of time and then that digital item is superimposed on the customer’s image later during the VTO. There are several techniques for creating these digital assets including various photogrammetry techniques where many images of the object are taken or scanned and then stitched together to create a textured model of the object. These include Apple’s new Capture SDK which was released just last month and many 3rd party groups. However, these techniques don’t work well for every type of object including eyewear, jewelry and anything that has a lot of shiny, translucent, or transparent attributes.  These digital assets oftentimes need to be built using 3D software in a manual or semi-manual process to get accurate geometry and textures. 
4. The fitting and positioning of the digital asset onto the reconstructed face/body part/body or landmark points
5. The occlusions are figured out based on the positioning of the digital asset with respect to reconstructed face/body part/body in 3D.
6. The digital asset is projected back into the single image, a set of recorded images or a video, or onto a live, or onto the real-time camera feed

Note there are some AI techniques that avoid the need for digital replicas but they are in early stages (see Predictions for VTO below).

Virtual Try-On Use Cases 

Virtual try-on  can solve two key questions for customers:

1. Which product should I buy?
2. Will this fit me?

Product Discovery

Product discovery is a common use case for VTO. A customer is trying to pick a lipstick color and can virtually try-on all the colors to help with their selection.  In some cases like Ditto’s VTO, anatomical information about the face is used to determine face shape so flattering products can be recommended.

Fit

Fit is the number one reason for product returns which as noted above can be detrimental to a retailers or brand’s bottom line and create an inconvenience for customers to need to send back the product. Some VTO solutions are not just showing the products on the customers but are able to accurately size the product so the customer can be sure what they are buying online fits them and they won’t have to mess with a return. There are other fit-oriented solutions (like TrueFit) that are gaining in popularity that recommend sizes without the use of VTO and just use manually entered information about your body and other brands/sizes that fit to suggest a size.

Virtual Try-On Benefits

There are several key benefits for adding VTO for both customers and retailers/brands.

Improved Customer Experience

Virtual try-on technology makes online shopping experiential. Customers can see themselves in products which helps with both engagement and confidence to make purchases. Information on which products are tried on can also be used to further personalize the experience.  Better customer experiences lead to increased loyalty and word-of-mouth – the best marketing channel available.

Improved Conversions and Average Order Value

Virtual try-on helps consumers decide which products to buy and can get more confident what they are buying fits them. This confidence leads to higher purchase frequency and higher average order value (AOV). 

Not every user will engage with the virtual try-on technology but for those who do, quality virtual try-on solutions see massive lift in conversions and AOVs.  For Ditto, we see a 2-5x lift in conversion rate when customers use our virtual try-on technology when implemented properly and a 7-9% increase in AOV.  This is a huge lift in sales for our clients.

Reduced Returns

As mentioned above, returns are a massive issue for eCommerce sites and up to 30-50% of purchases for fitting items get returned. Poor fit contributes to ⅔ of these returns. We see luxury, fitted items at the highest end of this range.

Some virtual try-on tools accurately size the object on the user which can help them pick out items that fit.  This can drastically reduce the return rate and mean huge savings for retailers and brands and fewer hassles for customers having to deal with returns.

Increased Personalization and Purchase Frequency

Virtual try-on is an opportunity for retailers to learn more about their users – what products they are interested in, what styles fit them, etc.  Retailers who are able to harness this data can personalize their site experience for each user and create much more relevant communications to that user post-visit which leads to an increased frequency of purchase.

Improved Store Experiences

While most people think of virtual try-on as an eCommerce feature, we are seeing very compelling use cases in physical retail stores.  These can occur on a shared device like a kiosk or on a personal device like a user’s smartphone often prompted by a QR code.  When used in a physical store, virtual try-on can be used in several ways:

1. One common way is to help customers through product discovery. We’ve seen this in make-up where customers can use their smartphones to virtually try on lipsticks before picking out a color in a store without physical sampling. 
2. We have also seen many “endless aisle” use cases – where customers can see themselves in merchandise that’s only available online. This is commonly used to show customers additional colorways or additional sizes. This can be key to help retailers move to a showroom model which has massive efficiencies on inventory management.  Retailers can keep “showroom” products on the floor and ship the final products from centralized distribution centers.
3. Virtual try-on can also be useful for categories that have tricky in-store physical try-on experiences. For example, many buyers of prescription glasses have a visual impairment. When they take off their glasses to try on a new pair, they often can’t see themselves in the mirror. Ditto’s virtual try-on can make it easy to see themselves in a new pair while wearing their existing pair of glasses.

Better Omnichannel Connections

The best retail experiences today typically include both eCommerce and stores seamlessly connected by technology.  Products virtually tried online can be pre-pulled and ready to physically try on when you go into the store.  eCommerce and store experiences can be personalized and streamlined for each user. 

Key Considerations When Selecting a VTO

Fit

Fitting the object onto the user accurately is a huge benefit of VTO that can materially increase conversions and reduce returns as discussed above.  There are a number of ways VTO solutions scale objects relative to the users. Popular methods include using standardized objects (usually a card from your wallet since it’s a universally sized object), manually-entered measurements, depth sensors on the device, and the use of computer vision to detect features alongside AI algorithms to make educated guesstimates on the size based on the detected feature. Some VTO solutions also make guesses based on gender. This is obviously problematic if the feature varies widely within the gender.

If you sell products where fit is a key criterion to the purchase, consider using a VTO provider with an accurate scale which can have a huge impact on your conversion rates and returns rate and has a much higher ROI than a VTO that makes everything look like it fits.

Mobile-Oriented

Having a VTO solution that works well on mobile is a no-brainer given consumer trends towards the use of mobile phones even to make eCommerce purchases. A few years ago the split of Ditto’s end users was about 50/50% desktop mobile. Now over 75% of our end customers use our tech on mobile devices.

It’s also a bit easier to use VTO on a mobile device because of the front-facing camera and the ability to adjust your lighting by moving around. Customers are also very familiar with selfies and navigating and optimizing their front-facing cameras.

Web Solutions vs. Native App

We want to highlight the distinction between mobile web users and mobile native app users. The latter requires customers to go to the App Store and download an application. 

This tends to be a fit for organizations with a high purchase frequency (think multicategory retailers like Amazon, Walmart, department stores, pharmacies, etc). 

We have seen a native application strategy be a problem for categories (like glasses) that aren’t purchased frequently because customers don’t want to download an application for a purchase they make 1-2 times a year. The result is a low percentage of downloads relative to the customer base. For these categories, a mobile-optimized web experience is the way to go. 

Native applications can have certain key benefits to the user experience if you can get over the download hurdle. For example, iPhone 10+ devices have a TrueDepth sensor that can make it much easier to build VTO solutions. There are a number of solutions limited only to iPhoneX+ for this reason.  Native applications also have an easier webcam and microphone access where users opt to let the application access to the camera, camera roll, microphone, audio, etc. onc. Then return users don’t have to deal with webcam access on subsequent visits. With web experiences, there are more frequent requests to access the webcam. Sites can use cookies to remember preferences for some period of time but eventually these cookies will clear and the site will need to get webcam and microphone access again.

Browser/Device Agnostic

You will also want to evaluate the browsers and devices your customer base uses so we recommend looking at your site analytics. As mentioned above, some VTO solutions only work on certain devices like iPhoneX+.  Other VTO solutions only work in some browsers. The Apple’s AR Kit works on the web in a limited fashion only on Safari (not Chrome or other browsers). Similarly, Google’s ChromeKit functionality only works on Chrome.  Many other web VTO solutions will only work on HTML5 browsers. While most modern browsers do support HTML5, older versions of the browsers don’t. This means many VTO solutions won’t work on older browsers at all. The browser windows within social media sites also historically struggled with HTML5 although this is changing now with the latest browsers accessible through Facebook/Instagram, etc.

The ability to access VTO from a wide range of browsers and devices easily directly correlates with the percentage of your users that will use the VTO. It’s also very challenging to get customers to change their browsers so we recommend making sure your VTO solution works where your customers are today.

Privacy and Consent

Anytime you are dealing with a customer’s face or data, it’s important to work with VTO vendors who can protect that data. Some VTO solutions are run locally and save no data. Other VTO solutions run server-side and process data in the cloud. If using a VTO provider that has any server-side processing, you’ll want to ensure they are compliant with all relevant laws.

Many times, this means getting the proper consent from the customers and being clear about what data you are processing, where the data is stored, how long you are storing data if at all, and notifying them about their rights to delete it if they’d like. VTO vendors need to be able to support this functionality.

As privacy laws get stronger country by country and state by state, VTO providers need to have a robust infrastructure to support in-country data storage, the ability to delete past users down the log file, dynamic and robust data storage capabilities. Ask about GDPR, AAPI, HIPAA compliance, and more.

High-Quality Digital Assets

For most VTO solutions today, there is a digital replica of the physical product that is created ahead of time – often called the digital asset. This digital asset is then superimposed on the customer’s image later during the VTO.  

There are many ways to create digital assets.  One popular technique is photogrammetry where many images of the object are taken or scanned and then stitched together to create a textured model of the object.   Apple’s new Capture SDK is a newer example of this and there are many 3rd parties who can take this imagery and create digital assets.  This does require the physical product to be present when scanning which can create logistical issues if you are trying to create digital assets from samples or if you have a large number of products.

Also worth noting, physical scanning and photogrammetry techniques don’t work well for every type of object including eyewear, jewelry and anything that has a lot of shiny, translucent, or transparent attributes.  These digital assets oftentimes need to be built by talented Digital Artists using 3D software like 3DMax, Maya, Blender, etc to replicate the mesh/geometry and the material/texture of the object.

As a general rule of thumb, the geometry of the digital asset can be used between VTO solutions since most 3D file formats can be converted into .obj files.  There are often adjustments needed (like the polygon count) to make it work for another VTO solution but there are ways to convert meshes amongst technologies.

However, the materials tend to be highly optimized for the specific rendering engine used. See here more information on rendering engines. 

As you might expect, the costs to produce high-quality digital assets vary drastically by category and the skill level required. Understanding the number of digital assets needed, the process needed to create digital assets, and the platform and rendering engine used are all key considerations when choosing a VTO solution.

Digital Assets Library

In addition to being able to make new digital assets at high quality, it’s important to understand if your VTO Solution already has a large library or database of Digital Assets you can use.  This can make the process and cost to add VTO significantly easier, faster, and with less experience.

As an example, Ditto has the world’s largest library of digital eyewear SKUs. This means our clients who sell designer eyewear don’t have to worry about digitizing products because most of what they sell is already created and they can use them off the shelf. For clients who design and manufacture their own products, we have a workstream that can digitize thousands of products a month.  

As an example, Ditto has an expansive database of digital glasses that we consistently update. We currently have over 110,000 glasses digitized and add thousands more each month.  If you need additional glasses digitized, our process is simple. Using your existing images and a single measurement, Ditto can do the work of creating digital glasses for you.  

The Realism of the Rendered Digital Asset  

The quality of the digital asset is huge but so is the rendering engine used to show that digital asset on the user.  There are some common rendering frameworks, some available on specific platforms, others working in servers as well as mobile devices. Which framework to choose depends on the materials and requirements of the product or on the requirements of the application; meaning the industry in focus. Some frameworks tend to work well on opaque, larger items like furniture. We’ve seen these work well for shoes, suitcases, etc too.

Some of these common frameworks are often used for very high-quality rendering but often not in real-time. Think of auto manufacturing sites that allow you to “build” your own car configuration. They pre-render the car configurations so you can build up your own.  Industries like film and television, games, architecture, etc also use rendering engines that can produce photorealistic results. Virtual try-on applications either live or video-based has minimal rendering time requirements since the user would like to interact with different products without a significant wait time.

For Virtual Try-On, the rendering engine has much more real-time performance implications because the Digital Asset needs to be rendered onto the user’s face, body, etc, in real-time or with just a few seconds of delay. This can be done client-side or server side but there isn’t time to pre-render outputs ahead of time. Speed and quality are often trade-offs so VTO tends to have a higher weight for speed while trying to achieve a certain level of realism.

For many categories, custom shaders or rendering engines are needed to make the products look realistic when being virtually tried on. This includes objects where there are a lot of reflective properties, translucency, and transparency including jewelry and eyewear. Many of the top VTO providers spend a lot of money creating customized rendering engines and shaders to up the ante on realism. The extent of custom shaders required also depends on the variety of products you would like to support. For example, if the product library only consists of limited materials or limited products with similar characteristics, one of the common frameworks or limited custom shaders can be the solution for rendering technology. However, if the product library consists of different materials (opaque, shiny, transparent/translucent, etc of a variety of products, more effort needs to be put into custom shaders. 

Product Recommendations and Personalization

As mentioned above, Product discovery is a common use case for VTO. 

For example, for a VTO for lipstick, there is the ability to show the various lipsticks on the user and then there is the further benefit of recommending colors for them based on skin tone, trends, past purchases, purchased from people who look like them, etc.

In some cases like Ditto’s VTO, anatomical information about the face is used to determine the customer’s face shape and measurements so several flattering styles can be recommended. The customer adds their personal style preferences and eyewear is recommended to simplify and accelerate finding the right pair.  

These insights about the customer and the data collected by tracking their behavior with the technology can lead to highly personalized user experiences and helpful product recommendations.  Faster and easier product discovery makes the buying process more enjoyable and leads to higher customer satisfaction, sales, and average order value.

These insights can also be used to personalize marketing communications. 

When customers save their Ditto virtual try-on, you have the opportunity to capture their email and build a relationship with them through your CRM tool. In addition, the Ditto Marketing API also allows you to send personalized emails that feature their face with any pair of glasses for them to try on. 

User Experience

Virtual try-on solutions can have a wide range of user experiences including real-time augmentation,  augmenting products on record images or videos, or even the use of an avatar to try-on products. 

When evaluating which user experience is best for your products, consider things like whether your product is bought on impulse or highly considered. If it’s an impulse buy, quick engagement VTO solutions might be a better choice. If it’s a highly considered purchase, precision of fit and the ability to save the Try-On so the customer can move between eCommerce and physical retail stores might be preferred.

Also consider any unique aspects of the try-on experience for your category. For example, 63% of Americans wear prescription glasses. The percentage of visitors who are actively shopping for glasses online is even higher.  As such, it’s important to factor in the customer’s visual impairment even if you are selling sunglasses.  Ditto’s unique recorded video user experience enables customers to wear their glasses and still see themselves on the screen.

There are also several avatar user experiences used mostly for apparel VTO where getting a high quality full body image is challenging. Customers can enter measurements to create their own virtual model or avatar and is then used to virtually try-on clothes.  There are new hardware solutions like Amazon’s Echo that are trying to make it easier to get a full body capture. 

Virtual Try-On Category Leaders

Virtual Try-On is becoming table stakes for many industries where the ability to try on the product is critical to online conversions. 

Make-up and Hair

Beauty is leading the way with several very compelling virtual try-on solutions.  Because there are no physical digital assets, VTO solutions are a little easier to build and are most dependent on great real-time face detection technology which is becoming pretty mature. This is leading to very high-quality real-time solutions for make-up. Hair detection is harder but improving and there are also several new real-time VTO solutions for hair.  Here are a few of the leaders in the beauty AR category:

1. Modiface 
   1. Modiface is the leader in beauty AR with very good facial feature detection and a strong user experience. L’oreal, the largest global cosmetics company, bought Modiface in 2018. In addition to creating strong native apps and browsers experiences for L’Oreal, Modiface licenses its technology to major players including Amazon and Sephora. It also just announced a partnership with Facebook to bring make-up virtual try-on to Instagram and Facebook. 
2. YouCam
   1. Perfect Corp. has built a variety of beauty apps under the YouCam brand for makeup, hair, nails as well as image and video filters and face editors. These solutions are used by beauty retailers and brands including Sally Beauty, Laura Mercier, and Madison Reed.  They also announced a partnership with Chinese retail giant Alibaba.
3. FaceCake
   1. FaceCake has a wide scope and creates native and web solutions from jewelry to apparel to accessories to makeup and home decor. We have seen their tech on beauty apps like MAC and Physicians Formula. They also provide AI-driven recommendations and fit assessments for some categories. 
4. JD
   1. Chinese retail giant JD has been building AR solutions in-house and released virtual try-on solutions for its native apps for beauty and shoes. 
5. Banuba
   1. Banuba has a wide scope and creates native and web solutions from beauty to eyewear to jewelry. They license their core tech so developers can build custom applications for everything from social media filters to shopping plug-ins.

Eyewear

Eyewear is another leading category for virtual try-on. Unlike beauty, eyewear VTO requires digital assets to fit on the customer’s face.  Yet it’s an easier VTO product than clothes because eyewear is more rigid and doesn’t require the complexity of fit like clothes with elasticity and drape and customers are comfortable with selfies.  It’s also a category with a huge need given the low online penetration of glasses and the importance of fit to the purchase.  

1. Ditto
   1. We believe we have built the leading VTO for eyewear globally. Our proprietary technology accurately detects the users face and allows us to create a very accurate 3D reconstruction of each user’s face. We create very accurate, high quality digital assets and have fitting algorithms that accurately place the glasses onto the face so customers can see how the glasses fit before they buy them.  Ditto has a unique video user experience that works well for customers who wear glasses and is device and browser agnostic. Ditto has a large library of digital assets (over 110K and counting) and is able to make product recommendations based on what will fit and reflect each customer’s personal style. Ditto’s solution is used by leading eyewear retailers and brands including Zenni, Specsavers, GrandVision, Misterspex, Quay, Ace & Tate and more. 
2. Fittingbox
   1. Fittingbox is the global leader in real-time web-based VTO for eyewear. They also have a large library of digital assets. You can read more about the comparison between Ditto and Fittingbox here. Fittingbox is used by Fielmann, Jins, and Maui Jim and more.
3. Warby Parker
   1. Warby Parker is an eyewear brand who has built their own virtual try-on app for iPhone X+ devices. It uses the TrueDepth sensor and allows customers to do a real-time virtual try-on of their products on their iOS app.

Shoes

1. Wanna
   1. Wanna (also known as Wannaby) has created some very cool augmented reality apps for real-time native app shoe virtual try-on. They license their technology to show powerhouses including Adidas, Nike, Gucci, Farfetch and more. They also have a Nails and Watches application that are solid. 
2. Snap
   1. Snap (previously Snapchat) has been building their own AR solutions (and buying auxiliary companies including FitAnalytics and Screenshop) to lure brands to create AR Ads on the platform. One of their more impressive use cases is their in-app shoe try-on. 
3. Vyking

1. Vyking’s SneakerKit is a real-time shoe virtual try-on that works on a browser. This is key for retailers and brands without a native app and they have landed some big accounts including Adidas and Under Armour.

Clothes

Virtual try-on for clothes is still in its infancy without many great real-world solutions live today. It’s a much tougher category to nail technically but the benefits will be massive to the winners.Given patents and whitepapers and early prototype launches, it’s safe to say that there will be some big launches from the largest retailers (Amazon, Alibaba, JD, etc.) and social platforms (Snap, Facebook/Instagram, etc) for apparel virtual try-on in the years to come. Here are a few we have our eye on:

1. Snap
   1. Snap (previously known as Snapchat)’s beauty and shoe try-on are just the beginning. They’ve recently announced some new AR capabilities for clothes and have partnered with key brands including Prada so we’ll keep an eye out for more developments.  
2. Amazon
   1. This is not a productized solution yet but Amazon is developing technologies with a GAN-based (Generative adversarial network) virtual try-on which is an AI technology that demonstrates on target image how a combination of products would look like. Amazon’s technology, called “Outfit-VITON” takes as input a query image and a number of reference garment images. The query image is an image of a person and the network generates a new image with the same person but wearing the reference garments. 
3. Alibaba
   1. Alibaba’s TMall has demoed in-store virtual try-on solutions for apparel with a specialized mirror so we suspect there is more coming from the Chinese retail giant.
4. Clo
   1. Clo Virtual Fashion is a leader in 3D software for apparel designers and has impressive clothing rendering technology. It’s not a virtual try-on solution but this level of realism for virtual clothes is leading the way.
5. Walmart
   1. Walmart recently bought Zeekit, a leading virtual fitting room platform that has a virtual try-on platform for apparel and a platform that uses augmented reality to pre-render and morph apparel onto pre-captured models of various body types.   We’ll see how they integrate that into their own sites and apps going forward.

Predictions on the Future of VTO

Virtual Try-On will become Table Stakes

As virtual try-on technology continues to evolve from gimmicky to growth stage and eCommerce trends continue on their current tear, we’ll see virtual try-on go mainstream. Beauty and eyewear are leading the way followed by shoes, jewelry and eventually apparel.

The Line will Continue to Blur between Social Media and eCommerce

Snap’s AR toolkit is becoming increasingly robust.  Facebook and Amazon are building a large suite of AR applications too.  Social media are creating features to make it easier to buy in-app and retailers are adding more social network capabilities. Massive players in China like Alibaba already do both and are paving the roadmap for how to bridge the gap. We expect all social media and retailers to both leverage the power of AR virtual try-on. 

Depth Sensors will Become Prevalent and Accessible by the Browser

Apple’s TrueDepth sensor that powers FaceID has led the pack on depth sensors in mobile devices. Android is quickly trying to catch-up and we expect the majority of high-end Android devices to have depth sensors of their own in the years to come.  

Today the full power of the TrueDepth sensor (face tracking, face/body pose estimation, 3D mesh generation) is only accessible in iOS native applications. Apple has started to release some aspects of the ARKit to their Safari web browser but it’s limited. We predict Apple will open up the full capabilities of the TrueDepth sensor to the browsers starting with Safari and eventually to all browsers. This will enable improved virtual try-on applications on the mobile web like being able to scale objects accurately without the need for a standardized object.

Improved Phones will Lead to an Uptick in Realism

New and powerful chips manufactured and included both in modern mobile devices and laptops open the path to more powerful technologies on the client side. One such technological solution is more realistic rendering through ray-tracing techniques. Today, the computational power needed to ray-track can only be done outside of real-time solutions. But as these new chips continue to improve, the processing time can be significantly improved and real-time solutions with impressive realism in colors, shadows, and materials can be produced. 

In addition to rendering technologies, new powerful chips also open the path to the integration of AI-based computer vision technologies such as face/body part/hand/hair detection, face/body mesh generation, facial landmark prediction, joint prediction, body/face/hand pose estimation. These technologies will enable more accurate 3D geometries, more accurate fit, and hence a more realistic try-on. 

AI Visualizations will become Productized

We expect more AI virtual try-on solutions using Generative adversarial networks (GANs). These solutions are image-based and different from other VTO solutions that also generate 3D shapes of the face/body/body part. Using a generative model and a discriminative model, they aim to generate realistic images with certain conditions, such as the inclusion of garments, adapting the colors based on a reference, changing the appearance of objects, etc. The generative model is responsible for updating the query image (image of a person, face, etc) whereas the discriminative model is responsible for determining fake images. Through an iterative learning process, the generative model learns to produce realistic images that would be approved by the discriminative model. Long story short, it’s a new method that could really open up virtual try-on for apparel especially if paired with a fit calculation.

There will be a handful of winners per category.

Virtual try-on has such compelling financial impacts for retailers that the race is one for the big tech players to develop proprietary solutions.  We suspect Amazon, Walmart, Alibaba, and JD will build their own virtual try-on solutions, negotiate exclusivity with partners or buy companies in the category so they can have a competitive advantage.   

Outside of the major platforms, we see a handful of winners emerging in each key category, largely driven specialized solutions that are expensive to replicate and huge libraries of digitized products and data that reduce friction for retailers and brands to add virtual try-on to their sites. 

Personalized omnichannel experiences will win 

The winners of brick-and-mortar retail will seamlessly integrate their online and store experiences using virtual try-on, product recommendations, and order ahead as key links between the two.  Virtual try-on will enable retailers to “showroom” products in their stores and ship the final products from centralized distribution centers. This will enable an “endless aisle” while creating massive efficiencies for inventory management. Personalization will continue to make both the online and store experiences relevant and communications from the retailers/brands will continue to get more relevant, interesting, and effective over time.

Summary

Virtual try-on is evolving into a powerful, engaging experience. Customers get key questions answered around product discovery and fit while experiencing the products in a fun new way. Retailers benefit from increased conversion, reduced return, and a better customer experience that drives loyalty.

The technology is starting to mature in beauty and eyewear with shoes close behind. Technology will continue to evolve and we are within striking distance of incredibly compelling virtual try-on solutions even for apparel.