Supporters of Marcus Endicott’s Patreon can access weekly or monthly video consultations on this topic.
Tuoyuan Wisdom, marketed internationally as X-Era AI and earlier branded ExpanderaAI, is a Sun Yat-sen University–rooted Chinese artificial intelligence company specializing in digital-human content generation, livestream automation, and, more recently, embodied intelligence. The principal legal entity, Tuoyuan (Guangzhou) Wisdom Technology Co., Ltd. (拓元（广州）智慧科技有限公司), was registered on 1 December 2021. The company is headquartered in the Nansha District of Guangzhou, with a Beijing branch and a separately registered Shenzhen subsidiary known as Tuoyuan Wisdom (Shenzhen) Technology Co., Ltd. The legal representative recorded in public business filings is Peng Jiefeng, while the founder and chief executive is Lin Liang. The company maintains two principal web properties, ex-ai.cn as the corporate site for X-Era AI and yuan-avatar.com as the dedicated marketing portal for the Yuan Avatar / Yuanfen Shen B-side platform, both operated by the same corporate group rather than by separate entities, with the production application running on the subdomain yuan.ex-ai.cn.
The company is identified as one of the first-batch ecosystem enterprises of Pengcheng Laboratory (鹏城实验室) in Shenzhen, a national-level artificial intelligence research platform whose computational infrastructure and policy alignment Tuoyuan has consistently emphasized. Lin Liang’s appointment in 2024 as director of Pengcheng Laboratory’s Multi-Agent and Embodied Intelligence Research Institute (鹏城实验室多智能体与具身智能研究所) deepened that institutional anchoring. Tuoyuan is registered among Guangdong province’s certified specialized, refined, distinctive, and innovative (专精特新) small and medium enterprises and is described in publicly edited reference materials as a state-level specialized and new artificial intelligence enterprise.
Lin Liang anchors the academic and technical narrative of the company. A professor of computer science at Sun Yat-sen University and founder of the SYSU Human-Cyber-Physical Intelligence Integration Laboratory, Lin was elected an IEEE Fellow in 2023, an IAPR Fellow in 2022, and an IET Fellow, and received the National Outstanding Youth Science Foundation grant in 2023. His research distinctions include the ACM NPAR 2010 Best Paper Award, the ICME 2014 Best Student Paper Award, a 2012 Google Faculty Award, an ICCV 2019 Best Paper Award nomination, the Wu Wenjun AI Natural Science Award, a Pattern Recognition journal Best Paper Award, and a first-prize award from the China Society of Image and Graphics. A postdoctoral period at the University of California, Los Angeles brought him into collaboration with Song-Chun Zhu. Between approximately 2016 and 2018 Lin held senior positions at SenseTime, described variously as Chief R&D Director, Executive Director of the SenseTime Research Institute, and Distinguished Scientist.
The chief scientist is Wang Guangrun, a former Oxford research fellow under Philip H. S. Torr and a top-tier inductee of Huawei’s Genius Youth Program (华为天才少年计划). The Tuoyuan Wisdom – Sun Yat-sen University Joint Laboratory is led on the academic side by Liang Xiaodan, an associate professor in the SYSU School of Intelligent Engineering, director of the General-purpose Embodied Intelligence Center, and a recipient of the Alibaba DAMO Academy Young Scholar (青橙) award. Other named senior researchers in company press materials include Wang Keze, identified as a national-level young talent and Wu Wenjun AI prize recipient, and Liu Yang. The co-founder most associated with the digital-human commercialization arm is Huang Weipeng, a Sun Yat-sen University computer-science alumnus and serial mobile-internet entrepreneur, who serves as general manager of the Yuanfen Shen platform. Huang previously participated in establishing the Shenzhen content platform Tuzhan (兔展智能) during his university years and describes himself as a ten-year mobile-internet practitioner.
Tuoyuan markets two principal consumer-facing brands on top of an underlying multimodal foundation-model stack the company calls its multimodal cognitive AI engine, branded internally as XEra-ULM for the unified language model, with vertical industry-specific models labelled xVLM. Yuanfen Shen (元分身), also rendered as Yuan Avatar, is the B-side digital-human service. Republished company materials describe Yuanfen Shen as a full-stack self-developed digital-human infrastructure provider delivering four service modules covering digital-human cloning, text-to-video creation, interactive livestreaming, and multimodal interaction. The stated production pipeline takes approximately ten minutes of recorded audio-video footage from a real subject, plus several days of back-end model training, to deliver a digital twin usable for unattended short-video output, with the higher-fidelity B-side pipeline taking up to seven days end-to-end. The platform supports multilingual text-to-speech, multiple voice timbres, background and subtitle editing, batch generation, and lip-synchronization that the company asserts is sufficient to pass casual Turing-style observation by non-specialists.
A July 2023 product update added livestream-specific functions described in company press releases as industry-firsts, including walking-while-broadcasting mobility, hand-held product demonstrations, GPT-driven question-and-answer replies, an Yuan takeover mode (元接管模式) allowing a human operator to step into the digital host’s stream, and multi-variable interactive guidance. The company asserts a two-week iteration cadence for the livestream stack. Fenshenbao (分身宝), the C-side mobile-application sibling of Yuanfen Shen, launched as a WeChat mini-program in late January and early February 2024 ahead of the Lunar New Year. According to company announcements, Fenshenbao requires thirty seconds of online recording to produce what the company describes as a mirror-grade digital twin in five minutes. The Lunar New Year campaign generated cross-platform exposure that the company has quantified, in unaudited self-issued figures, as more than four thousand first-day registered users, a peak share rate of synthesized greeting videos approaching three hundred percent, participation by close to two hundred self-media key opinion leaders, and roughly 1.4 million cross-platform impressions.
Underlying both consumer brands is a three-module architecture combining a foundation multimodal generative model, a causal model (因果模型) intended to make outputs physically plausible and stylistically controllable, and a mind model (心识模型) intended to capture user intent, taste, and value alignment. Wang Guangrun has publicly framed the architecture as a non-Transformer, non-mixture-of-experts neural design with dynamic adaptation in feature-vector space, engineered to reduce training and inference cost. By 2025 the company’s discourse pivoted toward physical-space intelligence with a Vision-World-Action (VWA) architecture positioned as a successor to the Vision-Language-Action paradigm in robotics, complemented by an embodied foundation model named RoBridge co-published in May 2025. While digital-human services remain the productized core, the corporate narrative has visibly shifted toward embodied artificial intelligence and physical-world reasoning, a trajectory underscored in December 2025 when the company demonstrated a high-precision indoor navigation system at the 19th China Information Accessibility Forum.
Three publicly reported financing rounds form the visible record. Angel and angel-plus rounds were closed in 2022, with some reports specifying December of that year, for a combined “tens of millions of yuan” (数千万元). The disclosed angel-stage investors are Ginkgo Valley Capital (银杏谷资本), Yuanshu Investment (源数投资), Zhuoyuan Capital (卓源资本), and Hanren Capital (汉仁资本). A Pre-A round close to one hundred million yuan (近亿元) was announced on or around 30 December 2024, with Yueke Financial Group (粤科金融集团), the Pengcheng Vision Fund (鹏城愿景基金), and the Hongniao Qihang Fund (红鸟启航基金) as disclosed investors and stated proceeds directed at multimodal foundation-model and embodied-intelligence research and development. A larger Pre-A series round of several hundred million yuan (数亿元) was announced on 9 December 2025. Among its participants, Detao Capital (德韬资本) is the industrial-investment vehicle linked to listed home-furnishings manufacturer Goldenhome Living (金牌家居, 603180.SH) and the Jianpan Group, with which Tuoyuan has subsequently announced collaboration on a product called Feiliu AI (飞流AI) for AI-assisted interior-design generation. Shixi Capital (石溪资本), founded by integrated-circuit memory leaders, also participated, alongside Shujucheng (数聚乘) and Lingyi Investment (领屹投资), with continued backing from the earlier Pre-A investor consortium. Strategic listed-company affiliates SigmaStar Technology (星宸科技) and Eastern Pacific Industry (东方精工, 002611.SZ) entered through related vehicles, and Eastern Pacific has confirmed equity participation through the Shenzhen Stock Exchange’s investor-interaction platform. The detailed cap table is not publicly disclosed in open sources.
The most concrete partnership cluster centers on Guangdong-area state and platform actors. As a portfolio company within the Pengcheng Laboratory compute ecosystem, Tuoyuan also cooperates with ByteDance’s Volcano Engine cloud platform, with Alibaba’s international business through Alibaba.com, and with the Taobao Creative Center, while customer-operated digital-human livestreams run on JD.com, Tmall, and Pinduoduo. In March 2024 the company opened a Nansha Digital-Human Livestream Base (南沙数智人直播基地) in cooperation with a Nansha District state-owned investment platform, focused on cross-border e-commerce livestream services for the Guangdong-Hong Kong-Macao Greater Bay Area. A Beijing Fengtai District municipal newspaper item dated 6 December 2023 noted Tuoyuan’s establishment of a presence in that district. Beyond these named arrangements, the company states an enterprise reference list exceeding one hundred leading brands, central and state-owned enterprises, foreign companies, and listed companies across healthcare, insurance, internet platforms, legal services, and consumer electronics, although very few names are public and verifiable case-study outcomes such as sales lift or conversion uplift are not disclosed.
The company’s research apparatus is a major point of marketing emphasis. Press releases cite more than ten Best Paper Awards from international conferences and journals, the China Society of Image and Graphics first-prize science-and-technology award, the Wu Wenjun AI Natural Science Award, a provincial Natural Science first prize, and more than thirty wins in international AI competitions, although many of these are awards held by the founder, the chief scientist, and the SYSU joint-lab principal rather than by the corporate entity. Tuoyuan’s involvement in standards setting is more clearly documented than its corporate patent portfolio. The company publicly states that in 2023 it co-released, jointly with the China Academy of Information and Communications Technology (CAICT, 中国信通院), the country’s first Trustworthy Virtual Human White Paper (可信虚拟人白皮书) and that it has acted as a core participating drafter of more than ten digital-human industry standards. The Yuanfen Shen platform is reported to have passed CAICT’s Trustworthy Virtual Human content-management system test and to have been the first Chinese company to pass CAICT’s digital-human virtual-livestream standard-conformance verification in 2024. It was also included in CAICT’s 2023 AIGC industry map.
China’s regulatory framework relevant to Tuoyuan’s products consists of several layered instruments. The 2022 Provisions on the Administration of Deep Synthesis Internet Information Services (互联网信息服务深度合成管理规定), which took effect in January 2023, require registration of deep-synthesis algorithms in the Cyberspace Administration of China algorithm registration system at beian.cac.gov.cn, prominent labeling of synthetic content, and various security obligations. The administration has continued to publish batches of the domestic deep-synthesis algorithm registration list, with the fourteenth batch released on 6 November 2025. Industry norms further include the National Radio and Television Administration’s GY/T 411-2024 standard on virtual-human services in broadcasting and an August 2025 Beijing Internet Court ruling that recognized copyright protection for virtual digital human content in specific circumstances. The newest instrument is the Cyberspace Administration’s draft Measures for the Administration of Digital Virtual Human Information Services (数字虚拟人信息服务管理办法), released for public comment on 3 April 2026. The draft defines a digital virtual human as a virtual digital figure existing in non-physical worlds, simulating human appearance with voice, behavior, interaction, or personality characteristics, and produced through graphics, digital image processing, or AI technologies, whether real-person-driven or compute-driven. It introduces a continuous digital-human labeling requirement on display interfaces and content distribution surfaces, mandates separate consent for processing of sensitive personal data used in modeling, prohibits unauthorized use of natural-person likeness or voice that is highly similar to a specific natural person, and prohibits services to minors involving virtual relatives, virtual companions, induced over-consumption, religious inducement, and other content harmful to minors. Administrative penalties range from RMB 10,000 to 200,000 for severe violations. The Measures are still in consultation, with no effective date specified, and they sit alongside an earlier December 2025 consultation on Interim Measures for the Administration of AI Anthropomorphic Interactive Services (人工智能拟人化互动服务管理暂行办法), which addresses emotional-companion artificial intelligence agents. Tuoyuan’s product line, particularly Fenshenbao’s celebrity- and family-likeness use cases and the Yuanfen Shen livestream-host system, sits squarely inside the perimeter of these instruments.
Within the Chinese AI digital-human segment, Tuoyuan competes alongside specialist players such as Silicon Intelligence (硅基智能) and Xmov / MoFa Technology (魔珐科技), platform-scale offerings including Baidu Xiling, Tencent Cloud’s Digital Human as a Service, NetEase Fuxi, iFlytek’s digital-human stack, and the avatar product within the same Volcano Engine cloud platform that hosts Tuoyuan, as well as operator-driven offerings from China Mobile, China Telecom, and China Unicom. On a verifiable basis the company is best understood as a mid-tier specialist focused on digital-human content tooling for short-video and livestream e-commerce, distinct from the very large platform players that dominate analyst rankings of the Chinese digital-human market. Specialized industry reports typically segment leaders into the platform incumbents, with Tuoyuan more often classed among emerging players. The company differentiates on four narratives, all of them company-issued positioning statements rather than independent rankings: deep coupling with academic talent at Sun Yat-sen University, embedding in the Pengcheng Laboratory national-team compute ecosystem, full-stack self-developed generation, and cognitive-AI integration with the digital-human “brain.”
The verifiable corporate timeline accordingly runs as follows. Throughout 2022 the company built its initial product in tandem with the post-rebranding metaverse hype cycle and the subsequent generative-AI inflection, with the digital-human video product line formally launched as an internal project in June 2022 and AI-driven real-person cloning services publicly demonstrated by August. The angel and angel-plus rounds closed around the end of 2022, and the Yuanfen Shen platform launched its first commercial offering in December of that year. In 2023 the company joined CAICT in releasing the Trustworthy Virtual Human White Paper, was included in CAICT’s AIGC industry map, and progressively rolled out livestream-specific features in July. Fenshenbao reached the consumer market in late January and early February 2024 ahead of the Lunar New Year, and a substantive interview with Huang Weipeng appeared in 21st Century Business Herald on 20 April 2024. RoBridge was co-published in May 2025, and across that year the corporate narrative pivoted decisively toward physical-space intelligence. The Pre-A financing of close to one hundred million yuan was announced on 30 December 2024, followed by the Pre-A series financing of several hundred million yuan on 9 December 2025, with the indoor navigation demonstration at the 19th China Information Accessibility Forum closing the year.
Taken together, Tuoyuan Wisdom is a Sun Yat-sen University–rooted, Pengcheng Laboratory–embedded artificial intelligence startup that commercialized digital-human video and livestreaming through a B-side platform from late 2022 and added a C-side application in early 2024, while progressively repositioning toward embodied intelligence over the course of 2025. Its founder anchors a strong academic and industrial pedigree, its chief scientist contributes a research-grade architectural narrative, and its investor base spans early venture capital, Guangdong provincial state finance, Pengcheng-affiliated achievement-conversion vehicles, and listed-company industrial capital in home furnishings, integrated-circuit vision, and intelligent equipment. The company’s most ambitious self-claims, including full-stack self-developed foundation models, minutes-scale cloning, industry-first livestream features, and category-leading testing certifications, are partly substantiated by participation in CAICT testing and standard-drafting programs and partly remain corporate self-positioning. Against a Chinese regulatory environment that is tightening notably around digital virtual humans, Tuoyuan operates close to the regulatory frontier, with a product line directly in scope of the deep-synthesis regime, the broadcasting standard, and the new draft Measures on Digital Virtual Human Information Services.