This AR startup wants to reimagine animation with machine learning

For decades, avid fans of sci-fi and superheroes have become accustomed to holographic characters popping out of devices or dashboards of spaceships. What was once merely the figment of an imagination is now becoming a reality with the help of rapid innovation in alternate reality technologies.

Among those working on this concept is Scanta, a Gurugram-based augmented reality (AR) startup that is on a mission to change the future of communication and animation.

After completing his MSc in International Business from Prague College in the Czech Republic five years ago, Chaitanya Hiremath returned to his hometown of Delhi. He was particularly interested in the business side of emerging technologies.

This led him to launch his first startup, Fitness Studios, with the idea of creating something similar to Pixar Animation Studios that has made popular animated films such as Toy Story and Finding Nemo. It was during this time that alternate reality technologies, especially virtual reality (VR), caught Hiremath’s eye. 

“We worked for about eight months doing research and trying to develop a VR video. In the end we could only come up with a 15-30 second video animation,” Hiremath told TechCircle.

After this, Hiremath continued to carry out market surveys of VR expansion in India and the US but discovered that the technology was only just maturing and VR hardware was an obstacle.

“I had created a custom CPU and then used a heavy suitcase to lug around my gear from once place to another,” Hiremath recalled.  

He then began exploring augmented reality (AR). Unlike VR, AR didn’t need heavy hardware and Hiremath also felt that the technology could be adopted faster owing to the smartphone boom.

In the middle of 2016, he incorporated Scanta and started working closely with Google on its AR project called Tango. This led Hiremath and his then 10-member team to create what they claim is the world’s first AR-based basketball game.

Soon after iPhone-maker Apple came out with its own library for AR-based app and content development called AR Kit. Hiremath started working with it and developed more games such as first-person shooter offering Polygoons.

It was also during this period that Hiremath started envisioning the future of communication. His first gamble was to re-imagine emojis and hence he created AR avatars -- characters that represent online users in games and forums -- that would offer users a new experience. The company currently has a library of 110 avatars which can be accessed via its Pikamoji app.

However, Hiremath was not satisfied with merely creating avatars. He wanted to breathe life into them. “We decided to make the avatars intelligent by powering it with artificial intelligence abilities so that it could understand voice commands or inputs from users and react accordingly,” he explained.

This project was started around eight months ago. Scanta’s existing 26-member team -- 20 in India and the rest in the US -- is currently running beta tests and expects to make it available via a new app later this year.

“We are one step closer to a world where artificial intelligence can select keywords from a real-time voice command, analyse it, and then finally construct an AR avatar animation accordingly,” Hiremath said.

Hiremath and his team are eyeing a market with huge potential. According to a Markets and Markets report, the AR market is expected to grow from $11.14 billion in 2018 to $60.55 billion by 2023, at a compound annual growth rate of 40.29%. 

Technology

Hiremath said that Scanta was using a mix of AR programming languages, natural language understanding and machine learning to breathe life into its avatars.

Once the machine receives a voice input or query from the user, it breaks this down and commences sentiment analysis, contextual analysis and behavioural analysis before sending out the avatar reaction.

In the backend, Scanta has created mathematical equations between stages of animation development that include shading, texturing, and rigging, among others.

“Once the neural network determines or understands the command, the animation aspect takes over,” he said.

The company has also been working with Generative Adversarial Networks (GANs). GAN is a type of machine learning system under which two neural networks are pitted against each other. This technique has been widely used to generate photographs that resemble real individuals.

“We have also been collaborating with OpenAI,” Hiremath said. OpenAI is a non-profit AI research organisation aimed at promoting the ‘friendly’ use of the technology.

Scanta is also working with Google’s AR software library AR Core and a Japanese simultaneous localisation and mapping (SLAM) platform company for speedier development.

Funding and revenue model

Scanta secured its first funding cheque after participating in a Shark Tank Showcase event in the US, with New York-based investment company SB Enterprises committing $250,000 in March last year

It received a second investment subsequently -- from Germany's Accent Capital Group -- after Scanta was admitted to an accelerator programme.

Although Scanta has a monthly cash burn rate of $30,000, Hiremath said that product development rather than revenue remains the main focus.

Scanta has what Hiremath calls a trilinear approach to generate income. This includes building a consumer usage model and then taking it back to huge conglomerates as a marketing platform.  

The company has also run marketing campaigns based on events or promotions at a mall in Gurugram.

“The marketing campaign with Coca-Cola proved that people are very much interested in the technology. Our goal was to ensure sales of 3,000 cans from one spot. We ended up doing a little over 14,000,” Hiremath said, adding that it enjoyed similar success with fast-food brand Wendy’s.

Scanta has been in talks with smartphone manufacturers to feature as a pre-loaded app. Hiremath also told TechCircle that the company was working with companies in the home automation and assistant-based speaker segment. Without providing further details, he said that Scanta could soon be experienced in the form of holographic images in the future via a well-known digital assistant.

Film animation is another segment where Hiremath foresees growth. While this tends to be a time-consuming and expensive process, Scanta’s model of combining AI and animation will enable designers to create models merely using their voice.

“We believe we can become pioneers in the 3D animation space owing to our first-mover advantage,” he said.