Making Millions on YouTube: The Rise of the Virtual Influencer

Derek welcomes listeners from Japan and introduces Jia Shen from AKA Virtual, who has deep expertise in AI-driven virtual economy. Jia introduces himself as originally from Chicago and San Francisco, but having lived in Japan for 15 years.

Jia shares his background starting startups since 2005, focusing on social, cultural, gaming and entertainment. He came to Japan to pursue the Japanese gaming and IP world, working on Pokémon games, and collaborating with Sega, Capcom, and Bandai Namco.

Jia discusses attending Tokyo Game Show for the third year, sharing how TGS has evolved over the last decade. He highlights the rise of Chinese game companies (now representing a quarter of floor space), the shift from Japanese-only games to global AAA titles, and the unique challenge of designing booths for both business and consumer audiences.

Jia explains the AI Witch and Fortune Teller demos showcased at TGS, emphasizing the importance of giving characters a specific reason to exist. He discusses DE-AI's two product types (information guidance vs entertainment), and shares examples from Osaka Expo, Happy Pea Land, and Izumo Taisha temple, highlighting how fortune-telling creates an expected, specialized experience with tangible takeaways.

Jia shares AKA Virtual's approach to handling poor network and background noise at events. He explains their two reliable interaction models: a big red button (lowest common denominator approach) and QR code-based mobile interaction, emphasizing the importance of designing for normal users rather than tech-savvy early adopters.

Jia explains AKA Virtual's choice to focus on Japanese anime-inspired characters rather than lifelike human avatars. He discusses lessons from VTuber days, explaining that cartoons are meant to be caricatures, and animated characters set proper expectations that AI assistants aren't human replacements. He emphasizes that animation is more powerful and believable than trying to replicate reality.

Jia explains DE-AI platform, built on AKA Virtual's animation engine foundation. He describes two types of AI assistants: functional vocational assistants (information guidance, navigation, checkout) and entertainment characters (cultural information, fortune-telling). He highlights multilingual capabilities crucial for Japan's 45 million annual tourists.

Jia discusses plans to expose DE-AI platform, explaining the challenge that everyone wants avatar builders. He details support for VRM and Live2D standards, and emphasizes infrastructure-agnostic approach. He explains their strategy of using fast Japanese LLM (Shisa) for sub-second responses while GPT processes, highlighting the importance of mixing AI infrastructures.

Jia explains AKA Virtual's three-layer business model: fundamental technology (3D character infrastructure, AI LLM infrastructure), product layer (IP company aspirations like Disney/Shonen Jump, plus SaaS services), and agency business. He discusses how the agency business emerged from 'dogfooding' their own product to understand content creation and popularity.

Jia shares AKA Virtual's expansion to Indonesia, representing their entire Southeast Asia business unit. He highlights managing JKT48 (spin-out of AKB48), being the most popular virtual influencers in Southeast Asia, and expanding beyond virtual characters to become a marketing agency for anything Japanese and gaming in Indonesia.

Jia showcases JKT48 concerts combining real and virtual performers, with 25,000 paid attendees. He shares VTuber scale statistics: 50 billion views annually, 6% of all global views, 60 out of top 20 channels are VTubers. He highlights examples like Pekora making $4 million annually, with peak streams generating $250,000, emphasizing VTubers are meaningful influencers worldwide.

Jia discusses VTubers being used for educational content and tech tutorials. He emphasizes the value of VTubing for creators who want to be judged by content and personality rather than appearance, highlighting how virtual characters enable great creators without exposing identity risks.

Jia clarifies that virtual idols and LLMs don't mix—VTubers are strictly human. He discusses LLMs as tools to augment content creation abilities, addressing VTuber burnout (average 3-year lifespan). He envisions mixed influencers: core hours human streaming, rest AI-powered, and explains the shift from 'virtual influencers' to 'virtual characters' like Mickey Mouse—IP that exists everywhere.

Jia discusses AI singing technology, noting voice and music models are great but slow. He explains AKA Virtual focuses on speaking because they want AI characters to feel real. He highlights challenges: Western models have terrible Japanese, Chinese models are better for Japanese but limited. He emphasizes the difficulty of replicating anime character voices and comedic nuances, requiring specialized programming.

Jia discusses AKA Virtual's need to spin up new agents daily and the importance of infrastructure. He explains they don't want to worry about plumbing—they need agents with brain, voice, response capabilities, and SLA. He highlights Agora's value, especially companies with 'Agora on a chip' enabling infrastructure-ready solutions that can be calibrated online, emphasizing infrastructure up to the agent level.

Jia discusses the future of virtual economy, highlighting Indonesia's massive e-commerce volume despite low spending per user. He explains young people don't care if influencers are real—they believe in value. He predicts a fundamental shift: as AI generation becomes too good to distinguish from reality, people will stop believing anything that looks real, instead valuing trusted entertainment systems and virtual subsystems.

Jia's advice to AI creators: start playing with tools all the time. There's so much power in your hands now. Derek thanks Jia for sharing insights about AKA Virtual's journey and vision for the next generation of virtual economy.