Michael Jung, DeepBrain AI

DeepBrain AI specializes in programs that render speech-to-LLM text commands for the automated generation of video avatars and detection of deepfake content. The products created by the platform are capable of assisting bank customers, training insurance agents and helping customers generate content at a fraction of the time and cost it organic video production with live actors would require.

What does DeepBrain AI really do?

We currently sell three products: AI Studios, AI Human and Deepfake Solutions. AI Studios allows users to create talking videos without filming a real person. With our solution, users simply input a script and generate a video. Our technology synthesizes voice and combines it with lip-syncing so the avatar’s lip movements match the audio.

AI Human integrates an avatar with a chatbot, allowing people to interact with a digital person. Our avatars appear on Shinhan Bank ATM screens. Samsung Life Insurance also uses AI Human for role-playing exercises, allowing sales agents to practice pitches with an avatar playing the customer.

Our third product, Deepfake Solutions, detects whether videos, audio and images are AI-generated or authentic. Today, we can clone a face, lip movements and voice using just 10 to 20 seconds of data. As fake prescriptions, mileage photos and accident images become easier to generate, demand for detection is growing. Our customers include the Korean National Police Agency.

What safeguards do you have to prevent people from abusing DeepBrain AI’s own technology?

Any individual can use AI Studios to create talking-avatar videos. We recognize the potential for misuse, so we have content moderation in place. For sensitive areas such as religion or politics, we ask users to our sales team before creating content.

For business-to-business customers, we provide more flexibility because large enterprises have their own reputational risks and responsibilities. Our contracts specify that we provide the technology, while customers are responsible for its use. Individual consumers, however, cannot create custom avatars of third parties. They can create only their own avatars.

How are you differentiating AI Human from similar platforms being developed by competitors?

Our main focus is making the experience more realistic. The key is overcoming the uncanny valley, a challenge also faced in robotics. We want to deliver better videos with less effort and fewer prompts, remaining competitive among top-tier companies. Around 80% of customers do not know whether they are interacting with an avatar or a human when speaking with a bank agent at a kiosk. However, the avatar is only one component of the solution.

Three components work together to deliver conversational AI. The first is speech-to-text, which converts speech into text. The second is the brain, powered by a large language model (LLM), such as ChatGPT or Claude, which generates answers. The third converts those answers into voice and delivers them through the avatar. Customers want to know whether the system understands their questions and provides accurate answers. Recognizing accents and transcribing speech, including children's speech, remain challenges. And, unlike traditional chatbots that rely on predefined questions and answers, LLMs enable more flexible conversations.

Expansion and sales are important when evaluating performance and sustainability, but I also look at whether customers continue using our solutions. Contract renewals are an important measure. Fortunately, we have not yet had a customer cancel a contract after adopting our technology because they continue to see its value. Our strategy is to remain visible and accessible when people need these solutions. We want potential customers to us so we can demonstrate use cases that fit their needs, supporting more organic business growth.

What do you think will remain for humans as AI avatars gain more ground in industries that previously used real actors and video creators?

Many people still do not use AI solutions, but once they understand the technology and see how it can improve their lives, adoption becomes difficult to avoid.

Many people still do not use AI solutions, but once they understand the technology and see how it can improve their lives, adoption becomes difficult to avoid.

Some offline assistant roles can be replaced by avatars and LLMs, while video production is another affected area. These changes are inevitable. As technology develops, the types of jobs people do will change. AI is another stage in historical transformation.

I do think around 80% of future entertainment characters could be AI-generated, while 20% of demand will remain for real people. This includes fictional anime characters and avatars based on real people. Some audiences will still want to meet real people in person, so both will coexist. One example of this coexistence is our work with Samsung Electronics. The company uses our platform to sell Galaxy phones on live-commerce shows, using an AI version of a well-known gold medal-winning skater. A traditional two-hour production of this nature would cost around $20,000, while our platform can reduces the cost to $100. The celebrity is paid separately by Samsung for the use of his intellectual property rights, while we provide the avatar technology and platform.

We are also seeing opportunities in role-playing and training. Insurance agents can practice conversations with virtual customers before dealing with real customers. We develop scenarios with different customer profiles and situations, allowing agents to practice responding. Beginners can follow scripts, while experienced agents can practice more naturally. Employees can develop their skills in a simulated environment before applying them in real situations. This is another way AI can support people rather than simply replace them.

What is your strategy for expanding global business development and building larger sales networks?

Adoption of new technology often depends on transition costs. If a business-to-business solution is significantly cheaper than the traditional approach, customers have an incentive to buy it. Producing a 10-minute video traditionally can cost $3,000 to $4,000, whereas our AI Studios subscription starts at around $10.

The same applies to AI Human. If a bank can use an avatar integrated with an LLM to handle customer interactions for a fraction of the cost of a human employee, there is a clear incentive to adopt it. As AI improves and cost savings become more apparent, more businesses will use these technologies.

What partnerships and international expansion strategies are you pursuing outside Korea?

Geographically, we initially target wealthier countries willing to invest in new technology before expanding into other markets. Over the past three to five years, the financial sector, particularly banking, has been a major customer. These companies have the resources to invest in new technology and improve their services.

Technology tends to move first, and policymakers then establish the necessary rules. As with automobiles and the internet, governments recognize technology's potential to contribute to economic development and seek a balance. We are also seeing AI legislation emerge in Korea and the European Union.

Lastly, our solutions are also not limited to one industry. If a technology works for one sector, we explore whether it can fit another. Like email, we believe these solutions can be universal.