DeepBrain AI (CEO Jang Se-young) was founded seven years ago in 2017. Even before the concept of generative artificial intelligence (AI) emerged, it established itself as a frontrunner within the digital human field based on image and voice synthesis technology.
There have been many major results which have grow to be topics of conversation each domestically and internationally. It was already 2019 when the present president was first introduced as a digital human. Since then, it has been creating issues by introducing AI announcers and historical figures.
And with the addition of generative AI, this field has recently been experiencing a full-blown boom. Firms like Syndicia have recently launched a business service that enables users to create AI avatars and create videos using mobile phones or webcams, and have begun to expand their business in earnest. Within the meantime, many related corporations at home and abroad have also increased.
Actually, DeepBrain AI owns a lot of the relevant technologies. The corporate is now specializing in developing business models in addition to technological advancements.
Lee Jeong-su, DeepBrain AI’s sales director, also said that trends have modified quite a bit over the past few years. “I actually have been working within the business for the reason that early days of generative AI, and I can feel that the present trends have modified significantly for the reason that early days of the technology,” he said.
The term digital human has modified to AI human. Within the early days of technology, the usual was to create a sensible avatar, that’s, one which looked as similar and natural as possible to an actual person. Subsequently, as a way to construct a digital human, a considerable amount of data needed to be collected through long-term filming or recording.
But things have modified recently. Practicality is the important thing.
It has also grow to be necessary to have the option to create content in a brief time period with only a small amount of knowledge and put it to use in various domains. The virtual influencer technology recently released by the worldwide platform ‘TikTok’ can be for producing short-form videos.
DeepBrainAI’s business portfolio is as diverse as its history. Representative examples include ▲’Rememory 2′, a memorial service for the deceased that began in 2022, ▲’Dream Avatar’, which creates your personal AI human with a single photo, and ▲’AI Studios’, which might produce videos based on generative AI technology.
‘Rememory 2’ is a video production service that uses photos and audio files of the deceased with the consent of the family to send messages. It became a hot topic when it first appeared, and demand has been increasing recently.
Dream Avatar, which creates a private AI human with just a photograph and a 10-second voice, has found proper demand within the insurance industry. It’s getting used as an ‘AI business card’ for workers.
AI Studios, which supports automatic content summarization and customised background and music, has expanded beyond B2C to B2B, with KB Insurance recently deciding to officially introduce it.
As well as, we’re providing AI Studios-based lecture production services that provide ‘AI professors’ to cyber universities, and we’re in discussions to supply Dream Avatar technology to the ‘Find Missing’ campaign.
Director Lee Jeong-su said that amongst them, he’s most concerned with the mix with commerce. It’s because “the food and beverage industry, including confectionery, is coming up with ideas on find out how to increase promotional effectiveness with AI technology.”
Essentially the most often discussed method is characterizing products. Using AI, products are personified, moved, and spoken, and appear in content. This has the advantage of giving stories to products.
We’re even considering a technique of getting real-time conversations with users by taking pictures of QR codes on the skin of product packaging. We’re pursuing collaborations with brands corresponding to Chips Ahoy and Pringles.
Last May, a patent application was filed for ‘deepfake detection’, a facial-behavioral pattern evaluation technology that considers not only a selected person’s face or appearance but additionally their behavioral patterns to find out the authenticity of a video.
This field can be the results of research and technological advancements which have been in place for the past 4-5 years. Director Lee said, “As we prepared countermeasures against the uncomfortable side effects of technology in keeping with the expansion of the business, it has grow to be a crucial item.”
To this end, we comprehensively analyze various behavioral patterns of the goal of imitation, corresponding to ▲head angle ▲lip utterance ▲facial muscle changes, through video data learning. After measuring the similarity between the video and the actual person, we offer the ultimate determination result. We plan to release a accomplished solution around September by combining it with the ‘stable diffusion’ detection function that reveals the source of the generated image.
As well as, we’re providing AI human kiosks to middle schools to support AI learning for youth. We’re also providing opportunities for planning, design, and video production using AI Studios.
Director Lee Jeong-su said, “The corporate’s principle is that business model development, prevention of uncomfortable side effects of technology, good use, and appropriate technology development should be intertwined.”
Lastly, he further explained the plan to advance AI human technology.
When evaluating the accuracy of AI humans, it is not uncommon to give attention to mouth shape and facial expressions. Nevertheless, DeepBrain AI goes beyond this and learns your entire facial and neck muscle movements into the model.
Here, they’re adding even color changes in response to the feel of the garments or changes in lighting. They said that they plan to release a hyper-realistic AI human that has learned all of this data after next yr.
We also plan to give attention to commercializing multi-devices to extend user contact points. We’re expanding to kiosks, mobile devices, and IPTV, and have accomplished development to permit greater than 1,000 users to attach concurrently.
He emphasized, “The market potential for generative AI and AI humans is wider than we predict,” and “For the reason that flow of related technologies and industries is so fast, we’re responding as flexibly as possible.”
Reporter Jang Se-min semim99@aitimes.com
