Three Decades of Virtual Human Development in Japan, China, and South Korea

The idea of the virtual human—a digitally constructed figure capable of performing cultural, commercial, or communicative functions once reserved for living people—has evolved over three decades from a curiosity of computer graphics into a globally significant industry. Three countries have done more than any others to shape this evolution: Japan, which pioneered the concept of the virtual idol and later gave rise to the VTuber phenomenon; China, which industrialized virtual human production at a scale unmatched anywhere in the world; and South Korea, which merged virtual character technology with its dominant entertainment export, K-pop. Each country’s trajectory reflects distinct cultural priorities, technological infrastructures, and regulatory philosophies, and the interplay among the three has produced a landscape of innovation, competition, and occasional geopolitical friction.

Japan’s claim to primacy in virtual human development rests on a lineage stretching back to the mid-1990s. In 1996, the major Japanese talent agency Horipro, in collaboration with Visual Science Laboratory, unveiled Kyoko Date, codenamed DK-96, a fully computer-generated three-dimensional idol whose development had begun the previous year. Kyoko Date is widely regarded as the first real-life virtual idol, distinguished from anime characters by her intended role as a media personality occupying the same promotional ecosystem as flesh-and-blood pop stars. She made her first radio appearance in October 1996 and released a CD single, “Love Communication,” the following month, but the project was a commercial failure, with the single selling fewer than thirty thousand copies amid widespread criticism of the uncanny valley effect her appearance produced. Shortly afterward, animator Kenichi Kutsugi created Terai Yuki, a three-dimensional virtual idol conceived in 1997 for his manga “Libido” and subsequently launched as a musical act through Avex Trax, with a CD single released on November 1, 1999. By 2000 Terai Yuki had appeared in television commercials and would go on to release four DVDs between 2000 and 2001. These early experiments, though commercially modest, established Japan as the birthplace of a concept that would prove extraordinarily durable.

The next transformative moment came with the development of Vocaloid, a singing voice synthesizer created by Yamaha Corporation in collaboration with the Music Technology Group at Pompeu Fabra University in Barcelona, under the leadership of Hideki Kenmochi. The software was first announced at Musikmesse in Frankfurt in March 2003 and officially launched on January 15, 2004, at the NAMM Show in the United States. Its first commercial voicebanks, LEON and LOLA, were English-language products released by the British company Zero-G Ltd., while the first Japanese voicebank, MEIKO, was released by Crypton Future Media on November 5, 2004. The technology’s cultural impact, however, remained limited until August 31, 2007, when Crypton released Hatsune Miku, designated CV01 in the Character Vocal Series, using the Vocaloid 2 engine. Miku’s voice was sampled from actress Saki Fujita and her character designed by illustrator KEI, but what distinguished her from earlier virtual performers was the participatory culture that erupted around her. Users on Nico Nico Douga created over a hundred thousand original songs using the software, and more than a million derivative artworks followed. Crypton actively encouraged this by launching the Piapro platform on December 3, 2007, and adopting a Creative Commons–influenced licensing framework, the Piapro Character License, that permitted non-commercial derivative works. Miku’s concert appearances further expanded the format’s reach. Her United States debut took place on July 2, 2011, at the Nokia Theatre L.A. LIVE, now the Peacock Theater, where a sold-out Mikunopolis show at Anime Expo drew an audience of thirty-five hundred. She returned to the same venue on October 11 and 12, 2014, for Miku Expo 2014, again selling out both performances. These concerts employed a rear-projection technique onto glass panels, sometimes described loosely as holographic but technically closer to the Pepper’s Ghost illusion, a distinction that became a point of controversy in 2024 when a touring production switched to LED screens.

The next seismic shift in Japan’s virtual human landscape arrived in late 2016 with the emergence of Kizuna AI, created by the Japanese technology company Activ8 Inc. under its Project A.I. initiative. Her YouTube channel, A.I.Channel, was established on October 18, 2016, and the first videos appeared on November 29 of that year. In a self-introduction video posted around December 1, 2016, she coined the term “Virtual YouTuber,” a label that would come to define an entire industry. Illustrated by En Morikura and modeled in three dimensions by Tomitake, with voice performance by Nozomi Kasuga, Kizuna AI popularized and named the VTuber format, although the first YouTuber to use a computer-generated avatar was Ami Yamato, who had begun doing so in 2011. What Kizuna AI ignited was a boom that generated an ecosystem of agencies, talent pipelines, and live-streaming platforms, the most prominent of which became Hololive Production, operated by Cover Corporation, and Nijisanji, operated by AnyColor Inc.

By the mid-2020s the Japanese VTuber industry had matured into a substantial commercial enterprise. Cover Corporation reported record revenue of approximately 43.4 billion yen, roughly 290 million dollars, for fiscal year 2025, with year-over-year growth of around sixty-nine percent in the third quarter. The company applied to move to the Tokyo Stock Exchange’s Prime Market, projected fiscal year 2026 revenue of 52.5 billion yen, and set a medium-term target of approximately one hundred billion yen by fiscal year 2030. Cover launched its Holoearth metaverse marketplace in April 2025 and an official card game, while also announcing a restructuring of its Holostars male VTuber branch, closing company-led activities as of April 3, 2026. AnyColor, the operator of Nijisanji, experienced a more turbulent period, with its stock declining by more than thirty percent through 2025 following talent controversies, though the company continued to grow revenue. Brave Group pursued aggressive expansion through acquisitions, purchasing Idol Corp in May 2024, followed by Neo-Porte and StelLive, while its subsidiary ENILIS acquired brossom. Avex spun off its virtual entertainment division into a new company called VEXZ in 2025. Perhaps the most striking development was the rise of Neuro-sama, an AI-driven VTuber operating on the Twitch channel of developer Vedal987, which by January 2026 had become the most subscribed streamer on the platform with approximately 162,459 active subscribers, surpassing every human streamer through near-continuous twenty-four-hour broadcasts. Estimates of the global VTuber market vary dramatically across research firms, with figures for 2024 ranging from approximately 1.5 billion to 3.87 billion dollars depending on the source and the breadth of activities included in the calculation, and compound annual growth rate projections spanning from under ten percent to over thirty percent.

Japan’s government responded to these developments with a series of policy and regulatory initiatives. In June 2022, the Intellectual Property Strategy Headquarters, operating under the Prime Minister’s Office, published the Intellectual Property Promotion Plan 2022, which identified legal challenges related to metaverse content among its top ten strategic measures, addressed the unauthorized imitation of real-world branded items in virtual spaces, and flagged the need for clearer trademark regulations governing metaverse products. Separately, the Ministry of Internal Affairs and Communications announced in July 2022 the formation of a Study Group on the Utilization of Metaverse Towards the Web3 Era, which held its first meeting on August 1, 2022, and produced a final report on July 18, 2023, after twelve sessions. The study group’s work covered avatar identity and harassment, digital infrastructure needs, democratic values in the metaverse, openness and interoperability, and international consensus formation. A successor body, the Study Group on Safe and Secure Metaverse, launched in October 2023 and produced its Principles of the Metaverse, Version 1.0, in October 2024.

China’s engagement with virtual human technology began in earnest in the mid-2000s, initially following models established abroad. HiPiHi, a three-dimensional virtual world platform comparable to Second Life, was formally founded in Beijing on April 17, 2006, after conceptual work that began in October 2005. The platform, which its CEO Xu Hui described as aimed at building a complete three-dimensional visual background reflecting different environments in the real world, entered alpha testing in December 2006, beta testing in March 2007 with approximately ten thousand users, and public beta in April 2008. HiPiHi partnered with IBM in 2008 but struggled with low engagement and ultimately shut down its servers in August 2012. The platform’s significance lay less in its commercial performance than in its demonstration that Chinese entrepreneurs were prepared to invest in virtual world infrastructure at a time when the concept remained speculative even in the West.

The scale of China’s subsequent commitment to virtual human technology has no parallel elsewhere. By 2024, according to data from Tianyancha, China was home to approximately 1.144 million companies involved in digital human–related work, with more than 174,000 new registrations in the first five months of that year alone. The core market was valued by iiMedia Research at 339.2 billion yuan in 2024, and projections from various sources pointed toward continued rapid growth, with IDC forecasting the enterprise-grade AI digital human market at 102.4 billion yuan by 2026. A city-level competitiveness index published in 2025 ranked Shenzhen first with a score of 92.92, followed by Beijing at 91.21, Guangzhou at 89.28, Shanghai at 86.36, and Hangzhou at 83.28. An earlier projection from QbitAI, published in December 2021 in its Virtual Digital Human Deep Industry Report, had estimated the total market at 270 billion yuan by 2030, split between identity-type virtual humans at approximately 175 billion yuan and service-type virtual humans at approximately 95 billion yuan. That estimate, once widely cited, has since been overtaken by far more aggressive figures, with iiMedia projecting the core market at 935.6 billion yuan by 2030 and the broader driven market at over 10.4 trillion yuan.

Several companies have emerged as leaders in China’s digital human ecosystem. XiaoIce, originally created by Microsoft as an AI chatbot in 2014 and spun off as an independent Chinese company in July 2020, has accumulated 660 million registered users globally and built a platform that has generated tens of millions of digital humans for applications including digital employees, voice cloning, and neural network rendering. Industry analyses by Sullivan and IDC have ranked XiaoIce in the first tier of China’s digital human market alongside Baidu and SenseTime. SenseTime, the artificial intelligence giant, launched its SenseAvatar platform on April 10, 2023, as part of its SenseNova large model ecosystem, offering AI-driven text generation combined with voice synthesis, lip-sync, and expression generation from as little as thirty seconds of footage, along with more than a hundred preset templates and over three hundred multilingual voices for applications including round-the-clock digital human livestreaming. SenseTime was ranked first in the IDC MarketScape evaluation for AI digital human products and participated in developing the first national standard for customer-service digital humans, designated GB/T 46483-2025 and published in November 2025. RM Group, the English name for Shanghai-based Ranmai Technology, carved out a distinct niche by creating AYAYI, widely recognized as China’s first hyper-realistic metahuman, who debuted on the social platform Xiaohongshu on May 20, 2021, and went viral. The company, co-founded by Nicky Yu, subsequently created additional metahuman characters including Noah and Lucy and developed sub-brands for offline and Web3 engagement. In the e-commerce sector, Silicon Intelligence had cloned more than five hundred thousand avatars and was running upward of fifty thousand daily livestreams.

China’s central government embedded virtual human development within its broader industrial planning apparatus. The Ministry of Industry and Information Technology, together with the Ministry of Education, the Ministry of Culture and Tourism, the State-owned Assets Supervision and Administration Commission, and the National Radio and Television Administration, jointly issued the Three-Year Action Plan for Innovative Development of the Metaverse Industry, covering the period 2023 to 2025, on August 29, 2023. The plan set a target of cultivating three to five globally influential ecosystem enterprises along with a cohort of specialized small and medium enterprises, and it explicitly called for the creation of well-known digital humans and benchmark digital human products. Beijing’s municipal government set its own target of building a fifty-billion-yuan digital human industry by 2025. At the standardization level, the MIIT’s Metaverse Standardization Working Group initiated work on virtual digital human classification, contributing to the broader effort to bring order to a rapidly expanding field.

The regulatory dimension of China’s approach has grown increasingly sophisticated. The Provisions on Administration of Deep Synthesis Internet Information Services, issued on November 25, 2022, by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the Ministry of Public Security, took effect on January 10, 2023, and imposed requirements for watermarking and labeling deep synthesis content, a category that encompasses digital human generation, face generation and replacement, and voice synthesis. On April 3, 2026, the CAC released a far more targeted draft regulation, the Digital Virtual Human Information Service Management Measures, for public comment. The draft defines digital virtual humans as virtual digital characters created using graphics, image processing, or artificial intelligence, and introduces mandatory labeling with the designation “digital human” throughout the entire display duration. It prohibits the use of virtual humans resembling specific real persons without consent, establishes special protections for digital recreations of deceased persons, bars the provision of virtual companion services to minors that could induce overconsumption, and requires real-name authentication. The draft envisions multi-agency oversight coordinated by the CAC in conjunction with telecom, police, cultural, health, financial, and broadcasting authorities.

South Korea’s contributions to virtual human development have been shaped decisively by the country’s preeminence in pop culture and entertainment technology. DeepBrain AI, headquartered in Gangnam-gu, Seoul, with additional offices in Palo Alto and Beijing, was founded in 2016 under the name MoneyBrain and developed AI Studios and AI Human, platforms for generating highly realistic virtual humans capable of real-time conversations and emotional gestures, with avatars developed from real people using their actual voices, physical appearances, gestures, and regional dialects. The company secured between 44 and 55.76 million dollars in total funding through a Series B round and deployed its technology across broadcasting, education, financial services, media, public services, retail, and healthcare, including an installation at Hyundai Department Store. In February 2026, as Korea’s AI Basic Act took effect, DeepBrain AI moved its deepfake detection capabilities into global infrastructure and joined the Ministry of Science and ICT’s digital innovation research and development project. Pulse 9, also based in Seoul and founded in late 2016 by CEO Park Ji-eun, developed its proprietary Deep Real AI engine for creating realistic virtual human faces from image data and a companion Deep Real Live system for live applications. The company created Eternity, South Korea’s first virtual K-pop girl group, an eleven-member act debuting in 2021, along with Zaein, a virtual human used for news broadcasting on SBS Morning Wide and for live commerce, serving clients including Shinsegae Group and L’Oréal Korea.

The most commercially significant South Korean experiment in virtual entertainment has been Plave, a five-member virtual boy band whose performers wear motion-capture suits, with their movements translated in real time into manhwa-style two-dimensional animated avatars rendered in Unreal Engine 5. The identities of the performers remain undisclosed, and the group writes, composes, and choreographs its own material. First introduced via livestream on September 15, 2022, and formally debuting on March 12, 2023, with the single album “Asterum,” Plave is managed by Vlast, a Seoul-based CG and visual effects studio turned intellectual property company led by CEO Lee Sung-gu. HYBE and YG Plus made equity investments in Vlast in April 2024. Plave’s commercial trajectory has been extraordinary: its third mini album, “Caligo Pt.1,” released on February 3, 2025, sold more than one million copies in its first week, making the group the first virtual idol act and the first boy group of 2025 to achieve million-seller status. Plave became the fastest act in history to surpass one billion streams on Melon, doing so in 494 days and breaking the record previously held by NewJeans, and reached two billion cumulative streams six days later. The group was the first virtual act to win first place on a music show, on Music Core, and the first virtual idol group to perform at the MAMA Awards in November 2024. A sold-out concert at Jamsil Indoor Stadium in October 2024 drew eleven thousand fans, and the group’s YouTube channel surpassed one million subscribers. A Japanese debut followed in mid-June 2025 with the single “Kakurenbo,” and in August 2025 Plave was appointed the first-ever Startup Seoul Ambassador. By contrast, Mave, a four-member virtual girl group created by Metaverse Entertainment, a subsidiary of Netmarble Corporation, in a joint development with Kakao Entertainment, took a different technological approach, using hyper-realistic three-dimensional CGI members created through machine learning and deepfake technology with MetaHuman and Unreal Engine. Debuting on January 25, 2023, Mave generated initial interest, with its debut music video accumulating over twenty-one million views, but the group did not achieve a commercial breakout comparable to Plave’s. By late 2024, multiple sources indicated that Mave was presumably disbanded or inactive, with no new content released since late 2023 or early 2024.

South Korea’s regulatory and policy infrastructure for artificial intelligence underwent a decisive transformation in late 2024 and into 2025. On December 26, 2024, the National Assembly passed the Framework Act on the Development of Artificial Intelligence and Establishment of Trust by a vote of 260 to 1 with 3 abstentions, making South Korea the second jurisdiction globally, after the European Union, to adopt comprehensive AI legislation. The law was promulgated on January 21, 2025, and took effect on January 22, 2026, after a one-year transition period. It adopts a risk-based framework, defining high-impact AI across eleven areas including healthcare, energy, nuclear applications, transport, recruitment, credit assessment, public decision-making, and education, though unlike the EU AI Act it does not designate outright prohibited AI practices. Transparency requirements include prior notification when using high-impact or generative AI, labeling of generative AI outputs, and explanation of AI outputs and criteria, with administrative fines of up to thirty million won, approximately 20,500 dollars, for violations. The government earmarked approximately 10.1 trillion won, roughly 6.94 billion dollars, for AI in its national budget, representing approximately a threefold increase, and five consortia led by Naver Cloud, SK Telecom, LG AI Research, NCSoft’s NC AI, and Upstage competed for 530 billion won in government funding for sovereign AI model development. Nvidia’s CEO announced a supply of 260,000 Blackwell GPUs for the Korean government and corporations in October 2025, and Ha Jung-woo of the Naver AI Innovation Center was appointed as South Korea’s first senior presidential secretary dedicated to AI on June 15, 2025, under the administration of President Lee Jae-myung, who took office on June 4, 2025.

A landmark legal ruling in South Korea tested the boundaries of how virtual identities interact with existing legal frameworks. On May 14, 2025, the Goyang Branch of the Uijeongbu District Court, in a civil damages suit brought by the five performers behind Plave through their agency Vlast against a social media user who had posted derogatory comments on X in July 2024, ruled that online insults directed at virtual avatars constitute defamation against the real individuals behind them. Judge Jang Yoo Jin reasoned that in an era defined by the metaverse, an avatar functions not as a simple virtual image but as a means of self-expression, identity, and social communication, and that defamation against an avatar can constitute an infringement on the actual user’s external honor. The defendant was ordered to pay one hundred thousand won to each of the five performers, totaling five hundred thousand won, a fraction of the 6.5 million won per performer that the plaintiffs had sought. The ruling does not grant independent legal personhood to virtual personas; rather, it extends existing defamation protections to the human beings who operate avatars, treating the avatar as an extension of human identity. Vlast appealed, seeking higher damages. The decision, made public on the court’s website in September 2025 and covered widely by media that month, established an early precedent for how legal systems might adapt to the growing cultural and economic significance of virtual characters.

The geopolitical dimensions of virtual human development surfaced most dramatically in the Hololive China incident of 2020, which illustrated how virtual performers operating across national boundaries can become entangled in political conflicts. In September 2020, Hololive VTubers Kiryu Coco and Akai Haato mentioned Taiwan while reading YouTube Analytics demographics data on livestreams, a seemingly innocuous act that provoked outrage among Chinese viewers on Bilibili because YouTube’s interface lists Taiwan as a separate region. Cover Corporation suspended both VTubers for three weeks on September 27, 2020, and issued statements that differed by language: the Chinese-language version posted on Bilibili explicitly supported the One-China principle, while the English and Japanese versions did not, a discrepancy that angered audiences on both sides. By October 22, 2020, Hololive’s Chinese VTuber Civia announced that all six members of the HoloCN branch, which had debuted in two generations on Bilibili between October 2019 and April 2020, were entering negotiations to end their affiliations. All six graduated between October and December 2020, with the final departure, Rosalyn, on December 27, effectively ending Hololive China. Harassment campaigns by Chinese nationalist groups against Kiryu Coco continued until her own graduation from Hololive on July 1, 2021.

Across all three countries, the virtual human industry has entered a phase characterized by institutional consolidation, regulatory maturation, and expanding commercial ambition. Japan’s strength lies in the depth of its creative ecosystem and the cultural legitimacy it has conferred on virtual performers over three decades, from Kyoko Date’s tentative debut to Cover Corporation’s trajectory toward a hundred-billion-yen business. China’s advantage is scale: more than a million registered companies, a national standards infrastructure, and a regulatory apparatus that, with the April 2026 draft measures, is moving toward the most comprehensive governance framework for digital virtual humans anywhere in the world. South Korea has demonstrated the power of integrating virtual human technology with an existing cultural export juggernaut, producing in Plave a virtual act whose commercial performance rivals that of the most successful human idol groups. The three countries’ trajectories are not merely parallel but increasingly interconnected, shaped by shared technologies, overlapping markets, and the recognition that virtual humans are no longer a novelty but a permanent and consequential feature of the global digital economy.

[Apr 2026]