Devar uses generative AI to create 3D AR images from text prompts

Tech company Devar is getting ready to launch the world’s first generative AI neural network for AR that allows people to create 3D objects and others AR assets.

The company’s Generative AR Platform brings together neural networks and cloud solutions to create AR content, including MyWebAR by Devar, a service launched in 2021 for creating AR content on the web. Together, it lets people create 3D AR images with simple text prompts. Once you create the 3D image, you can use it in an AR application on iOS or Android.

The first release of the platform will be a neural network for generating 3D objects and images optimized for performance, making them suitable for use in AR projects even on low-end and budget smartphones. The neural network allows users to create 3D and 2D objects for webAR projects, generating content non-stop until you have what you want.

“We are one of the few companies that has a completely proprietary computer vision and generative AI technology stack for augmented reality,” said Ana Belova, CEO of Devar, in an interview with GamesBeat. “We recently published, for example, text-to-3D — generative creation of ready-made 3D models by text. “Today we’ll start to see how people will use it in real life.”

In a demo, Belova (a Forbes 30 under 30 winner) showed how easy it was to create a 3D image and then put into an AR scene, like inserting a couch based on Vincent Van Gogh paintings into a room. You can create things like an AR business card to share with others.

“Our next level is generative AR platform which means you can create 3D models only using text prompts,” Belova said.

Devar’s own 3D content datasets and feedback from users of the MyWebAR platform were used to develop the neural network, which instantly generates an image designed for optimal tracking performance with the image recognition engine. Users can register on the MyWebAR platform to access the neural network.

“Professional studios usually have no problem with creating new characters for AR – they have 3D artists, designers and animators on staff to do this,” said “But according to the data of our no code platform MyWebAR, 60% of users have no experience creating AR. One of the main issues that becomes a barrier to their entry into the industry is the creation of 3D assets.”

“Professional studios usually have no problem with creating new characters for AR – they have 3D artists, designers and animators on staff to do this. But according to the data of our no code platform MyWebAR, 60% of users have no experience creating AR,” said Andrei Komissarov, Devar’s founder and CTO, in a statement. “One of the main issues that becomes a barrier to their entry into the industry is the creation of 3D assets. You need a lot of content, a lot of original requests. So we combined our data — our own library with thousands of objects, that we formed during 8 years of work in augmented reality, we took into account user requests and developed our own neural network that allows users to create 3D and 2D objects for webAR projects. Generate content non-stop — until you get the best result for you and your AR users.”

Origins

Devar has been toiling in the AR industry since 2015 with her CTO Komissarov. They have created one of the most successful mass-market AR products on the consumer market, with over 13 million Devar-branded AR books sold worldwide.

With those books, the company uses your smartphone camera to bring animated imagery to life when kids point it at a page of the book.

Sales of these kinds of books are nine times higher than kids books that don’t have AR capability, Belova said. Other applications include AR postcards that you can send to friends and have them see funny AR animations.

In 2021, the company launched its MyWebAR platform as a kind of Canva for AR projects. It helped people create AR without any need to know how to code. It can run on low-end laptops and Chromebooks.

“Two years ago, we launched a platform that allows you to create AR products without any technical skills,” she said. “Five years ago, you needed to be a skilled programmer to create mobile apps. Today, you don’t need to be a programmer. Our platform is like Canva.”

The company’s MyWebAR platform is one of the three most popular webAR platforms in the world, with brands and companies such as Google, Bath & Body Works, Hasbro, Dove, Colgate, Pandora using it, Belova said. These kinds of AR applications have enabled Devar to assemble more than 100,000 creators.

Each of these projects helped the company build its computer vision skills.

“Our camera can recognize an object and we have trained it over time,” she said.

And it turns out that those skills are what it needs in building AR experiences in the era of generative AI.

What makes text-to-3D-AR different?

Belova said that generated 3D models are optimized for performance, in order to make projects in AR work even on the low-end and budget smartphones.

The quality requirements for webAR objects include — the recommended number of polygons, correct topology, and the presence of textures. Usually it takes 3D specialists several hours or even days to create such assets. With cloud storage, AR content and software is stored on Devar’s servers, so you don’t have to worry about hosting.

It’s free for non-commercial use, with up to five AR experiences using the no-code editor. For $25 a month, creators can generate content with AI and get other pro benefits such as analytics. There are also Ultimate and Enterprise versions.

Images generated by the neural network can be used in augmented reality projects both as AR content and as a marker. The neural network instantly generates an image designed for optimal tracking performance with the image recognition engine.

The company has about 50 people and it’s based in the Bay Area. Belova hopes the market will take off with generative AR, but she knows the applications have to prove themselves.

“We have to understand that we have to educate the market,” she said. “The main problem has been the market. It’s not about technology and science. It’s also about how we use it to create and changing patterns of behavior.”

Companies like Niantic pioneered AR in games like Pokemon Go, but the AR features were often left turned off. The AR experience was too slow or not as visually impressive. It was a sign that people were looking for something better.

The AR future

Still, she’s excited about the AR future.

“It’s about how we can make the world around us really more helpful and useful. I give an example I mentioned. In the future, you have AR glasses, and you will sit on a crosswalk, and you would like to go,. bBut you will get notification some car around the corner isn’t stopping. This is about your safety.”

One day, Belova hopes the hardware experience will get better so we can see AR with higher-quality devices like the Apple Vision Pro. Or we could read books comfortably using comfortable AR glasses.

Belova said she would get the Apple Vision Pro and try to figure out what popular applications it will be able to run.

“I can’t wait when we have it in the market,” she said. “Imagine that you have AR glasses and you work not just at home but also out on the street and other places. You can read AR books. You can change your world. You can create something or collaborate. But we will have to all educate the market.”