DeepBrain AI (CEO Jang Se-young) was founded seven years ago in 2017. Before the concept of generative artificial intelligence (AI) officially emerged, it had already established itself as a pioneer in the digital human field based on video and voice synthesis technology.
During this time, it has introduced many notable results both domestically and internationally. Presenting a digital human of the current president for the first time was already a topic back in 2019. Following this, AI announcers and historical figures were introduced, creating significant buzz.
With the addition of generative AI, this field has recently experienced a full-scale boom. Companies like Synthesia recently launched a business service that allows users to create AI avatars and videos using mobile phones or webcams, expanding their business significantly. Many related companies have also increased domestically and internationally.
In fact, DeepBrain AI possesses most of the relevant technology. The company is now focusing not only on technological advancement but also on developing business models.
Lee Jung-soo, Director of Sales at DeepBrain AI, mentioned that trends have changed significantly over the past few years. "Having conducted business since the early days of generative AI, we can feel that the trends from the early stages of technology emergence and now are quite different," he said.
The term digital human has changed to AI human. Initially, the technology focused on creating avatars that looked and moved as realistically as possible, similar to real people. Hence, building digital humans required long hours of filming or recording to collect large amounts of data.
However, things have changed recently. Above all, practicality is key.
It has become important to create content with minimal data in a short amount of time and apply it to various domains. The virtual influencer technology recently released by the global platform TikTok is also aimed at creating short-form videos.
DeepBrain AI's business portfolio is as diverse as its history. Representative examples include ▲ the memorial service 'Re;memory started in 2022 ▲ 'Dream Avatar,' which creates a personal AI human with just one photo ▲ 'AI Studios,' which allows video production based on generative AI technology.
'Re;memory' is a service that creates videos using a photo and voice file of the deceased, with family consent, to generate messages. It went beyond being a novelty at its debut and is now seeing increasing demand.
Dream Avatar, which creates a personal AI human with just one photo and about 10 seconds of voice, found proper demand in the insurance industry. It is used as an 'AI business card' for employees.
AI Studios, which supports automatic content summarization and customized backgrounds and music, has expanded from B2C to B2B, recently being officially adopted by KB Insurance.
Additionally, they are providing AI professor services for cyber universities based on AI Studios for lecture production and are in discussions to offer Dream Avatar technology for a 'Missing Child' campaign.
Lee Jung-soo emphasized the integration with commerce. "The food and beverage industry, including confectionery, is brainstorming ways to enhance promotional effects using AI technology."
Characterizing products is the most discussed method. The idea is to anthropomorphize products using AI, making them move and speak, and appear in content. This can add a storyline to the products.
They are also considering a method where users scan a QR code on the product packaging to have real-time conversations. Collaborations with brands like Chips Ahoy and Pringles are in progress.
Last May, they filed a patent for 'Deepfake Detection,' which analyzes facial-behavior patterns. This technology determines the authenticity of videos by considering not only the face or appearance of a particular person but also their behavior patterns.
This field has also been researched and technologically refined over the past 4-5 years. Lee said, "As we expanded our business, we prepared measures to prevent the side effects of technology, which has become an important item."
To this end, they comprehensively analyze various behavior patterns of the target, such as ▲ head angles ▲ lip movements ▲ facial muscle changes, through video data learning. They measure the similarity between the video and the actual person and provide final determination results. They plan to launch a complete solution by September, combining it with 'Stable Diffusion' detection, which reveals the source of generated images.
In addition, they are providing AI human kiosks to middle schools, supporting AI learning for adolescents. They are also offering opportunities for planning, designing, and producing videos using AI Studios.
Lee said, "The principle of our company is that the development of business models, prevention of technological side effects, and appropriate technological development should be aligned."
Finally, he further explained the plan for advancing AI human technology.
When assessing the precision of AI humans, it is common to focus on mouth movements and facial expressions. However, DeepBrain AI is going beyond this by training models on the movements of the entire face and neck muscles.
They are also adding data on the texture of clothing and color changes due to lighting variations. They plan to launch a hyper-realistic AI human with all these data trained after next year.
Additionally, to increase user engagement, they plan to commercialize it across multiple devices. They are expanding to mobile and IPTV, in addition to kiosks, and have completed development for over 1,000 users to access simultaneously.
He emphasized, "The potential market applications for generative AI and AI humans are more diverse than expected," and added, "Given the rapid pace of technology and industry trends, we are responding as flexibly as possible."
