China's Digital Human Industry Comes of Age

The financial sector moved in parallel with state media. In 2019, Shanghai Pudong Development Bank partnered with Baidu Intelligent Cloud to develop Xiao Pu, described as the banking industry's first digital employee, with the concept announced in April, demonstrated publicly at the Baidu AI Developer Conference in July, and formally launched on December 13. The same year, in January 2019, the CCTV Network Spring Gala introduced digital twin hosts co-presenting alongside their human counterparts, with the digital twin of host Sa Beining, named Xiao Xiao Sa, created by ObEN, a US-based AI company. Subsequent landmark debuts populated the early 2020s. The virtual beauty influencer Liu Yexi launched on Douyin on October 31, 2021, by Hangzhou-based Chuangyi Technology, gaining over one million followers within twenty-four hours; AYAYI debuted in May 2021 from Ranmai Technology and became a digital brand manager for Alibaba's Tmall; and Feng Xiaoshu, modeled on the real presenter Feng Shu and built using Xiaoice's deep neural network rendering, was first deployed by Huafeng Weather Media Group during the 2022 Beijing Winter Olympics. NetEase's Fuxi Lab, established in 2017 as China's first dedicated game AI research institution, applied virtual human technology to the MMORPG Justice Online and integrated DeepSeek into the same title in February 2025.

Market sizing for the industry varies dramatically depending on methodology, and the gap between the most-cited estimates is itself a defining feature of the sector. IDC, applying a narrow definition limited to AI-driven digital human platform and SaaS revenue, reported the 2024 market at 41.2 billion yuan, representing 85.3 percent year-over-year growth, and projected it to reach 250.5 billion yuan by 2029 at a compound annual growth rate of 43.5 percent. The trajectory of these forecasts is itself instructive: IDC's June 2022 China AI Digital Human Market Status and Opportunity Analysis projected the sector would reach only 10.24 billion yuan by 2026, a target the actual 2024 figure already exceeded by a factor of four. IDC's June 2025 market share assessment placed Baidu first at 9.8 percent, followed by Huawei Cloud at 9.7 percent, Xiaoice at 5.1 percent, and SenseTime at 4.3 percent, with the 2D digital human sub-segment alone reaching 28.9 billion yuan in 2024, up 101.2 percent year over year.

iiMedia Research, by contrast, applies a substantially broader definition that captures the full ecosystem of content production, virtual commerce, virtual influencers, and downstream services. Its core market figures rose from 120.8 billion yuan in 2022, a 94.2 percent year-on-year increase, to 205.2 billion yuan in 2023 and 339.2 billion yuan in 2024, with a 2025 projection of 480.6 billion yuan and a 2030 projection of 935.6 billion yuan. The corresponding peripheral or driven market reached 4,785.3 billion yuan in 2024 and is projected at 10,468.6 billion yuan by 2030. The China Academy of Information and Communications Technology reported the industry scale at over 300 billion yuan in 2024 with growth above 35 percent, the China Internet Association projected the core market would surpass 400 billion yuan by 2025, Frost and Sullivan valued the AI digital human agent sub-market at 4.12 billion yuan in 2024, and Grand View Research projected a compound annual growth rate exceeding 53 percent for China's digital avatar market through 2030. An IDC platform-revenue estimate and an iiMedia full-ecosystem estimate for the same year can therefore differ by close to an order of magnitude, and any single figure cited as the market size should be read against the methodology behind it.

The structural metrics around the industry are less ambiguous. By the end of 2024, China hosted approximately 1.359 million digital-human-related enterprises, up 36.9 percent year-over-year, with over 413,000 new registrations during 2024 alone. Investment activity totaled thirty-three deals worth 110.56 billion yuan in 2024, patent applications reached 641 with 340 granted, an increase of 129.73 percent over 2023, and twenty-one provincial-level governments and forty municipal-level governments had collectively issued approximately three hundred specialized policies touching on the sector. By the same period, more than half of surveyed Chinese enterprises reported having used virtual human technology, a finding consistent across multiple industry surveys.

The single most important driver of the industry's acceleration has been generative AI. The integration of large language models into digital human systems beginning in 2023 collapsed production costs, shortened creation timelines from weeks to minutes, and enabled real-time interactive capabilities that were previously impossible. JD.com's Yanxi platform compressed single digital human production costs from tens of thousands of yuan to double digits, a reduction exceeding 90 percent. Zero-sample technology introduced in 2024 enabled instant digital human cloning from as little as thirty seconds of image data and ten seconds of voice data, while large language model inference costs dropped approximately sixty-fold across major Chinese cloud providers in a single year. Cloud rendering costs fell from approximately eighteen yuan per minute to roughly three yuan per minute, SaaS platform tools brought entry costs for small and medium businesses down from approximately five hundred thousand yuan for custom projects to twenty thousand yuan per year, and Baidu's Xi Ling 4.0, released in September 2024, reduced the entry price for 3D ultra-realistic digital human creation to 199 yuan. Generative AI technologies compressed development cycles from approximately three months to two weeks and reduced production costs by an estimated seventy percent overall, with one industry estimate projecting that generative AI will contribute over sixty percent of value in China's digital human market by 2027. The IDC MarketScape: China AI Digital Human Products 2024 Vendor Assessment, published in August 2024, evaluated sixteen vendors including Baidu, Mobvoi, Fengping Intelligence, Silicon Intelligence, Huawei, JD.com, MoFa (Shanghai) Information Technology, iFlytek, SenseTime, TRS, Tencent, Xiaoice, AsiaInfo Technologies, Zhongke Huilian, Zhongke Shenzhi, and Zhuiyi Technology, identifying AIGC digital humans, AI video generation, and multimodal large models as the top investment priorities over the next one to three years.

Text-to-speech advances have been particularly consequential for the sector. OpenAI's GPT-4o, announced May 13, 2024, introduced a single end-to-end multimodal neural network that reduced voice response latency from 5.4 seconds to as low as 232 milliseconds and enabled tone, emotion, and background-noise detection alongside expressive responses. Chinese companies responded within weeks. ByteDance published its Seed-TTS paper on June 4, 2024, then upgraded to Doubao-Seed-TTS 2.0 in October 2025 with enhanced emotional expressiveness; iFlytek launched its Her system as China's first ultra-fast super-human-like end-to-end voice-to-voice modeling system; MiniMax's Speech-02 achieved the top global ranking on the Artificial Analysis TTS benchmark with thirty-two-language support and ninety-nine percent speaker similarity in zero-shot voice cloning; Alibaba's CosyVoice 2.0, released in December 2024, achieved 150-millisecond streaming latency with MOS scores of 5.53 and was fully open-sourced under Apache 2.0; and Fish Audio's S2 Pro, a four-billion-parameter model released in March 2026, achieved the highest score on EmergentTTS-Eval, surpassing ElevenLabs, OpenAI, and Google. The open-source ecosystem proved particularly robust, with ChatTTS, F5-TTS, GLM-TTS from Zhipu AI, IndexTTS, and GPT-SoVITS all emerging across 2024 and 2025 and making high-quality Chinese TTS widely accessible.

A complementary wave of open-source digital human releases reshaped accessibility across the field. Silicon Intelligence open-sourced HeyGem.ai in March 2025, generating a digital clone from one second of video and a single photograph in thirty seconds before producing a sixty-second 4K video; the project accumulated over thirteen thousand GitHub stars. Alibaba's DAMO Academy released EchoMimic V2 for half-body animation, Tencent and Zhejiang University jointly released Sonic, JD.com open-sourced JoyHallo for Mandarin-optimized digital human generation, ByteDance released its LatentSync lip-sync framework alongside OmniHuman-1 in February 2025, and Ant Group's multimodal research lab open-sourced EchoMimic. The standard technical architecture now follows a pipeline of speech recognition, large language model reasoning and response generation, text-to-speech synthesis, and talking-head animation for lip synchronization, with retrieval augmented generation commonly used to supplement large language models with enterprise-specific knowledge bases.

The technology classification framework most commonly applied to the industry is the L1 through L5 automation hierarchy developed by the SenseTime Intelligence Research Institute and published in a white paper co-produced with the CARA alliance in April 2022. The framework ranges from L1, denoting CG-based static imagery, through L2 for motion-capture-driven systems, L3 for data and algorithm-driven systems with limited real-time interaction, L4 for real-time AI-generated figures with contextual understanding, and L5 for fully autonomous digital humans with self-learning capability. SenseTime's research institute also identified three defining characteristics of virtual digital humans: multimodal interaction, deep learning capabilities, and AIGC productivity.

The vertical structure of the industry is conventionally divided into upstream technology providers, midstream platform and application companies, and downstream deployment domains. Upstream providers supply foundational technologies in computer graphics, natural language processing, speech synthesis, computer vision, and motion capture; representative companies include iFlytek, with a full-stack AI virtual human platform offering more than sixty emotional voice libraries; Baidu, whose Xi Ling platform integrates its ERNIE large language model; SenseTime, which contributed to customer-service digital human classification standards; and Noitom, a leading Chinese motion capture hardware company. Rendering infrastructure relies substantially on foreign engines, particularly Epic Games' Unreal Engine and Unity Technologies' platform, though Chinese cloud rendering providers including Haima Cloud and Alibaba Cloud have developed real-time rendering services that reduce end-device hardware requirements. Midstream companies operate end-to-end platforms: Baidu's Xi Ling provides one-stop creation combined with LLM dialogue, Tencent's Zhiying offers SaaS-based video creation, ByteDance's Volcengine division runs interactive, broadcast, and livestream digital human products, JD Cloud's Yanxi serves e-commerce virtual livestreamers, and Kuaishou's Nuwa platform at peak capacity supported over 2,200 simultaneous digital human livestreams operating around the clock. In the virtual idol segment, Bilibili acquired the team behind Luo Tianyi, China's most prominent virtual singer, who was launched in 2012 and performed at the 2022 Beijing Winter Olympics opening ceremony.

Specialized vertical companies have established defensible positions in particular segments. Silicon Intelligence, founded in 2017 in Nanjing by Sima Huapeng, held 32.2 percent of the Chinese digital human agent market in 2024 by revenue, ranking first domestically and second globally, and filed for a Hong Kong IPO on October 31, 2025, positioning itself as the potential first publicly listed digital human company. The company completed eight rounds of funding totaling approximately 870 million yuan, with Tencent as its largest external investor at 16.59 percent ownership alongside Sequoia China, China Merchants Bank International, and other institutional backers; revenue grew from 223 million yuan in 2022 to 655 million yuan in 2024, with the company achieving its first adjusted net profit of 5.29 million yuan in the first half of 2025. Its financial profile reveals tensions characteristic of the industry: the top customer contributed 64.4 percent of revenue in the first half of 2025, and gross margins declined from 38.5 percent in 2022 to 31.6 percent in the first half of 2025 as competitive pricing strategies aimed at securing large enterprise accounts compressed unit economics. MoFa (Shanghai) Information Technology Co., Ltd., founded in 2017, completed its C round in April 2022 at approximately 110 million US dollars led by SoftBank Vision Fund 2, with combined B and C round funding of approximately 130 million US dollars from Sequoia China, 5Y Capital, and Northern Light Venture Capital, and serves more than two hundred enterprise clients including Alibaba, Tencent, CCTV, L'Oréal, and ByteDance. Xiaoice, spun off from Microsoft in July 2020 under the chairmanship of Harry Shum, operates more than three hundred thousand digital employees and completed a funding round of approximately one billion yuan in November 2022.

Sector deployments have advanced furthest in finance and media. The China Banking Association reported in November 2023 that eleven banks had deployed virtual digital humans, with five additional institutions in construction; named deployments include SPDB's Xiao Pu, which by 2022 was handling more than eighty percent of call volume and contributing to over three thousand person-years of digital labor across more than ten employee types and twelve channels, alongside CCB's Long Zhiwei and Long Zhiyuan, ICBC's Gong Xiaozheng and Gong Xiaocheng deployed in July 2023, and corresponding programs at Bank of Communications, Postal Savings Bank of China, China Everbright Bank, China Merchants Bank, China Minsheng Bank, Ping An Bank, Hangzhou Bank, and Changsha Bank. Combined technology investment by the six state-owned banks rose from 716.76 billion yuan in 2019 to 1,254.59 billion yuan in 2024, providing the budgetary substrate for sustained virtual human deployment. ICBC became the first major state-owned bank to deploy virtual digital human services on its mobile app, while Ping An separately deployed AI video consultants on digital screens across approximately thirty-four branches beginning August 2019, achieving a seventy-five percent AI replacement rate for certain customer service functions.

The media sector represents the most mature application domain. Beyond the Xinhua AI anchor lineage, CCTV developed Xiao C, the first digital human to conduct a live interview with a National People's Congress delegate, accumulating over two million social media followers, while the AI Wang Guan digital twin of CCTV financial commentator Wang Guan generated more than 780 million plays providing around-the-clock financial coverage. Multiple provincial broadcasters deployed their own digital anchors, including Hunan Satellite TV, Zhejiang Satellite TV, and Beijing Radio and Television, the last of which developed Shijian Xiaoni, a multi-functional digital human serving across news broadcasting, government Q&A, and the 12345 citizen service hotline. State-owned telecommunications operators have all deployed digital human solutions, with China Mobile launching its Lingxi AI intelligent agent in October 2024 based on its proprietary Jiutian large model.

E-commerce has been the most commercially visible and most contested application of digital human technology. JD.com's Caixiao Dongge, modeled on founder Liu Qiangdong, launched April 16, 2024, attracted over twenty million viewers in under an hour, and generated over fifty million yuan in transactions with user dwell time reaching 5.6 times the daily average. During the 618 festival in 2024, JD digital humans appeared in over five thousand brand livestream rooms with cumulative duration exceeding four hundred thousand hours and more than one hundred million viewers; by 2025, the platform rebranded as JoyAI reported cumulative viewing of seventeen million and gross merchandise value exceeding 700 million yuan, with digital human livestream costs at one-tenth of real-person costs. Baidu partnered with tech entrepreneur Luo Yonghao for a digital human livestream in June 2025 in which Luo's digital twin generated 55 million yuan in GMV from over thirteen million viewers in six hours, supported by an AI-generated script of 97,000 words and 8,300 AI-driven actions. Baidu's Huibao Xing platform, launched in October 2023, accumulated more than one hundred thousand digital human livestreamers, reporting conversion rate improvements of thirty-one percent and launch cost reductions of eighty percent. Platform regulation has varied sharply: Douyin banned fully AI-driven digital human livestreaming in May 2023, requiring real-time human oversight; WeChat Video Channels effectively classified virtual livestreaming as low-quality content in June 2024; Kuaishou withdrew traffic support for digital human streams the same month; and Taobao maintained a more permissive approach with vendor certification requirements. Zhejiang Province issued the first provincial-level guidelines for AI digital human livestreaming in September 2024, while sixty to seventy percent of digital human service providers active in 2023 had disappeared by 2024 amid intense consolidation.

Tourism and cultural heritage have emerged as a significant growth vertical. The Dunhuang Mogao Caves deployed Jiayao with Tencent on the Cloud Tour Dunhuang WeChat mini-program; the National Museum of China launched Ai Wenwen on July 22, 2022, the museum's 110th anniversary, developed with the Tencent SSV Digital Culture Lab, with the system after three upgrades covering over 1.4 million artifacts and providing guided tours for the Chinese Civilization Cloud Exhibition. The Capital Museum in Beijing introduced Jing Hui in September 2024 as an AI digital human guide for ancient Beijing history; the Chengdu Du Fu Thatched Cottage Museum deployed AI Du Fu, a digital recreation of the Tang Dynasty poet enabling visitor conversations about poetry, in September 2024; the Changchun Puppet Emperor's Palace Museum deployed Guide No. 5 with Beijing Bowen Technology in the same month; and the Datong Museum deployed an AI digital human guide for the Northern Wei Empress Dowager Feng exhibition in January 2025, developed by Acoustiguide Digital Culture, which had reached more than fifty museums by January 2025 with nearly one hundred additional partnerships confirmed. The 19th Asian Games in Hangzhou, held September 23 to October 8, 2023, served as a major showcase: the Digital Torch Bearer, known as Nongchaoer, enabled over one hundred million participants from more than 130 countries to create 3D digital avatars via an Alipay mini-program, and at the opening ceremony a giant digital human composed of these avatars co-lit the main torch alongside swimmer Wang Shun on a 185-by-20-meter screen, marking the first-ever digital torch-lighting in Asian Games or Olympic history. The system was developed by Alipay and Ant Group using a Web3D engine called Galacean, with Alibaba Cloud separately deploying the digital sign language interpreter Xiaomo for the Asian Para Games using a dataset of twenty-five thousand signs, and China News Service launching its AI anchor Xin Mei to deliver daily broadcasts and medal highlights. Translation services at competition venues were handled by approximately two thousand iFlytek hardware devices supporting eighty-four languages. China Unicom's hyperrealistic Liu Sanjie digital tourism ambassador, originally created on Baidu's Xi Ling platform and enhanced by Unicom Online, became the first provincial-level hyperrealistic tourism digital ambassador when launched for Guangxi province in March 2023.

The technology has surfaced new ethical and legal challenges as it has matured. During the 2024 Qingming Festival, the business of AI resurrection of deceased individuals surged, with over 1,900 merchants appearing on e-commerce platforms offering services priced from a few yuan to roughly twenty thousand yuan, and families of deceased celebrities including CoCo Lee, Godfrey Gao, and Qiao Renliang protested unauthorized AI recreations under Article 994 of the Civil Code, which establishes protections for deceased personality rights. In Hong Kong, police dismantled multiple syndicates using AI face-swapping for romance-investment scams, arresting twenty-seven people in October 2024 with losses exceeding 34 million Hong Kong dollars in one operation and thirty-one people in January 2025 with losses of 46 million US dollars in another.

China's regulatory framework for digital humans has developed incrementally over four years and reached a milestone in April 2026. The foundational layer is the Provisions on the Administration of Deep Synthesis Internet Information Services, effective January 10, 2023, jointly issued by the Cyberspace Administration of China, the Ministry of Industry and Information Technology, and the Ministry of Public Security, which explicitly covers face generation, face replacement, voice synthesis, and immersive realistic scenes, requires watermarks and labels on deep-synthesis content, and mandates separate individual consent for facial or voice biometric editing. The Interim Measures for the Management of Generative AI Services followed on August 15, 2023, applying to all generative AI services in China under a principle of balancing development with security through classification-based supervision. The Measures for the Labeling of AI-Generated Content, published in March 2025 and effective September 1, 2025, introduced mandatory national standard GB 45438-2025, requiring both visible AI-generated markers and machine-readable metadata embedded in files for all AI-generated content. On December 27, 2025, the Cyberspace Administration of China published the Interim Measures for the Management of Human-like Interactive AI Services for public comment with a January 25, 2026 deadline, regulating AI companions and emotional chatbots through provisions including a two-hour continuous use limit, mandatory break reminders, prohibitions on content encouraging self-harm or emotional manipulation, and special protections for minors and the elderly.

The most consequential regulatory development arrived on April 3, 2026, when the Cyberspace Administration of China published the Administrative Measures for Digital Virtual Human Information Services for public comment, with the comment period running through May 6, 2026. This twenty-seven-article regulation across five chapters represents the world's first attempt to create a dedicated legal framework specifically for digital virtual humans. Key provisions include a mandatory continuous digital human label throughout any digital human service display, a requirement for separate informed consent when using individuals' sensitive personal information for modeling or image generation with deletion rights upon withdrawal, prohibitions on creating identifiable digital humans of specific persons without consent, bans on providing virtual family members or virtual intimate partners to minors, prohibitions on inducement of excessive consumption, requirements that providers monitor user emotions and dependency and intervene in cases of extreme emotion or addiction, prohibitions on using digital humans to bypass facial or voice recognition systems, and graduated penalties from ten thousand to one hundred thousand yuan for standard violations and one hundred thousand to two hundred thousand yuan where violations endanger life or health. The Cyberspace Administration of China leads enforcement alongside more than ten participating agencies including the Ministry of Industry and Information Technology, the Ministry of Public Security, the Ministry of Culture and Tourism, the Ministry of Health, the State Administration for Market Regulation, and authorities responsible for financial regulation, broadcasting, publishing, film, and copyright.

At the local level, Beijing led with the Action Plan for Promoting Digital Human Industry Innovation and Development covering 2022 through 2025, issued in August 2022 as China's first dedicated digital human industry policy at any level of government, targeting fifty billion yuan in industry scale by 2025. The Beijing Digital Human Base in Chaoyang District opened in February 2024 and had attracted forty-nine enterprises by November of that year. Shanghai issued its own Metaverse New Track Action Plan in July 2022, including a Digital Human Comprehensive Enhancement Project as one of eight key initiatives, while Sichuan province issued its Metaverse Industry Development Action Plan in September 2023, specifically promoting functional, service-type digital humans across consulting, education, entertainment, and healthcare.

Industry standards have crystallized in parallel with the regulatory framework. The YD/T 4393 series, drafted by the China Academy of Information and Communications Technology under the China Communications Standards Association framework with co-drafting from Tencent, Baidu, NetEase, and SenseTime, was issued in 2023 and addresses appearance, voice, motion, interaction processing, and multimodal dimensions. GB/T 46483-2025, the first national standard for virtual digital humans, was led by SenseTime and published in November 2025, specifying minimum 200,000 polygons for 3D ultra-realistic models and 90 percent or greater lip-sync accuracy; the National Radio and Television Administration published its broadcasting industry standard GY/T 411-2024 in November 2024; and the Ministry of Industry and Information Technology released a mandatory national standard in January 2026 establishing a one-person-one-code digital human identity system requiring all commercial or communicative digital humans to carry unique identifiers for traceability. By December 2025, 748 generative AI services had completed Cyberspace Administration filing at the national level, and in February 2026 the Cyberspace Administration ordered removal of over thirteen thousand accounts and 543,000 pieces of content across Weibo, Douyin, Kuaishou, Bilibili, and WeChat for failure to label AI-generated material.

Among publicly listed Chinese companies in the space, several merit attention. Chinese Online operates the interactive visual novel platform Chapters, the My Escape platform, and the short-drama platform ReelShort. Alpha Entertainment holds animation IP including Pleasant Goat and Big Big Wolf and Super Wings, and in 2023 partnered with Xiaoice to create virtual digital humans for these properties through the joint venture Guangzhou Aochuang Huozhong Technology. Huace Film and Television has expanded into AI through its subsidiary Hangzhou Xinsuanyuanjie Technology, while Mango Super Media operates Mango TV under Hunan Broadcasting System. Yaowang Technology signed a strategic agreement with Xiaoice on May 15, 2023 to develop AI-powered livestream e-commerce. Wondershare operates the digital human video platform Virbo, known in Chinese as Wanxing Bobao, with multilingual templates targeting overseas markets in more than 120 languages. BlueFocus launched the virtual human Su Xiaomei on January 1, 2022 through its Blue Universe subsidiary and has also created the virtual musician K. SenseTime, listed on the Hong Kong Stock Exchange, underwent a leadership transition following the death of founder Tang Xiao'ou in December 2023 and produced a digital recreation of him in March 2024 through its Ruying AI 2.0 platform. iFlytek, which won a first prize in the 2024 National Science and Technology Progress Award, released its Spark Super-Realistic Digital Human in October 2024 as a system claiming semantically coherent lip-expression-action generation in a super-realistic format.

The trajectory of the industry as a whole is best characterized as a structural bifurcation rather than uniform growth or decline. Visual realism has advanced far ahead of conversational naturalness: industry insiders rated digital human interaction quality at only twenty to thirty points out of one hundred despite visual appearance scores reaching seventy to eighty points. People's Daily noted in September 2025 that the industry was stepping down from its idol pedestal and deepening focus on industrial needs, and MoFa CEO Chai Jinxiang told Jiemian News that many companies were eliminated because their digital human capabilities could not keep pace with large language model advances. The famous virtual influencer Liu Yexi required millions of yuan in production investment and hundreds of thousands per short video, and as of mid-2022 her parent company had not yet turned a profit despite a massive following; UBS estimated the average upfront cost of an advanced virtual idol at roughly thirty million yuan. Demand-side data from QuestMobile published in September 2025 painted a more cautious picture, reporting that virtual livestreamer gross merchandise value remained less than one-fifth that of real-person livestreamers in 2023, that average viewing duration on Douyin dropped from fifteen minutes in 2022 to five minutes in 2023, and that virtual streamer fan attrition rates exceeded forty percent on the platform. Yet the cooling narrative is only half the story: enterprise registrations hit a historical high of 413,000 in 2024, investment remained active at 110.56 billion yuan, and patent applications grew 129.73 percent year over year.

Consumer awareness is high but spending remains modest, and engagement is concentrated. Industry surveys found that more than fifty percent of surveyed enterprises had used virtual human technology, that 81.4 percent of respondents had encountered virtual livestreamers, that 92.3 percent of virtual idol fans were aged nineteen to thirty, that more than eighty percent of consumers spent less than one thousand yuan per month on virtual idol purchases, and that more than eighty percent indicated virtual idol endorsement would increase purchase intent. The iiMedia 2025 Digital Human City Development Index ranked Shenzhen first at 92.92, followed by Beijing at 91.21, Guangzhou at 89.28, Shanghai at 86.36, and Hangzhou at 83.28, with Wuxi in sixth position, distinguished by its manufacturing base and scenario deployment capabilities within the Yangtze River Delta region.

The strengths of the industry are well documented: rapid market expansion confirmed by multiple research firms, a deepening policy support infrastructure, a digitally literate consumer base of approximately 1.28 billion social media identities, a national internet user base that grew from 1.092 billion in December 2023 to 1.123 billion by June 2025, and a commercial ecosystem of more than 1,200 substantive companies with proven deployments across every major sector. The weaknesses are equally concrete: interaction capabilities, while dramatically improved by large language model integration, remain below human-level emotional depth and spontaneity; high-end custom 3D digital humans with professional motion capture still cost upward of one hundred thousand yuan; dependence on foreign rendering engines and the impact of US chip export controls on training compute availability represent structural vulnerabilities; and cross-disciplinary talent combining AI, computer graphics, natural language processing, and motion capture expertise remains scarce. Threats include the rapidly evolving regulatory landscape and competition from overseas providers including Microsoft, NVIDIA, Meta, Synthesia, and Soul Machines, with North America holding approximately 43 percent of the global market according to Mordor Intelligence. China's broader AI large model market reached 294 billion yuan in 2024, more than doubling year-on-year, providing the substrate within which the digital human sector operates. Underlying hardware development continued to advance, with Huawei's Ascend 910 series chips serving as a de facto standard for Chinese AI inference, Cambricon's Siyuan 590 entering volume production in the third quarter of 2024, and China Unicom and Huawei deploying 5.5G pilot networks in Beijing in January 2024 achieving ten gigabit-per-second downlink speeds capable of supporting real-time extended reality applications. The emerging deployment distribution settled into roughly half media humans, thirty percent service humans, and twenty percent industry-specific humans, a taxonomy reflecting both the technology's broadcasting origins and its expanding reach into commerce, government, education, and cultural tourism.

China's virtual digital human industry stands at an inflection point defined by the April 2026 draft regulation, which signals that Beijing views digital virtual humans as sufficiently consequential to warrant sector-specific governance rather than reliance on broader AI rules. The companies best positioned for the next phase are those with proprietary large language model integration, established enterprise relationships in regulated industries, or unique technical capabilities in 3D production and multimodal interaction. The industry's maturation will ultimately be measured not by aggregate market size figures that vary by hundreds of billions depending on scope definition, but by whether interactive AI digital humans can demonstrate sustained consumer engagement, measurable enterprise value, and responsible deployment within the governance framework that China is now constructing.

[May 2026]