Frost & Sullivan, in collaboration with LeadLeo Research Institute, hereby releases the annual report '2022 China Digital Human Market Observation Report - Milestones', part of the China Digital Human Series. The report aims to analyze the current development status, product characteristics, technological trends and commercial value of the Chinese digital human market, as well as to assess the competitive landscape of the Chinese digital human market and reflect the differentiated competitive advantages of leading brands in this market.
In the second quarter of 2022, Frost & Sullivan, in collaboration with LeadLeo Research Institute, conducted a downstream user experience survey on core products in the digital human field. Respondents came from various sectors such as internet, media, finance, with varying organizational sizes and specialized fields. The digital human trend analysis provided in this market report also reflects the overall trends of the digital human industry. The market rankings and leaders' judgments in the report are only applicable to the development cycle of China's digital human industry for the current year.
Digital humans refer to complex entities that exist in the digital world, created and utilized through computer graphics, graphical rendering, motion capture, deep learning, speech synthesis, and other computational methods. They possess multiple human characteristics (appearance features, human performance abilities, human interaction capabilities, etc.).
Based on geometric dimension characteristics, digital humans can be divided into 2D digital humans and 3D digital humans. 2D digital humans are limited to single-eye viewing, while 3D digital humans can achieve three dimensions in geometry through any visual viewing. 3D digital humans require the use of additional 3D modeling technology to generate digital images, which increases the information dimensionality and necessitates more computational resources. With the development of technology and real-world needs, 3D digital humans will become mainstream.
Data source: Frost & Sullivan consulting, LeadLeo research institute
Relying on the progress of interactive media, digital human-related technologies will continue to develop in directions characterized by strong immersion, interactivity, and openness.
Data source: Frost & Sullivan consulting, LeadLeo research institute
The core of the evolution of digital human production technology lies in the development of artificial intelligence technology. Relying on the progress of interactive media, future digital human-related technologies will aim to create digital humans with three-dimensional panoramas, realistic five senses, and hyper-realistic/digital twins. Key breakthroughs will be achieved through action and emotional interaction, as well as emotion perception and reasoning, continuously developing towards a direction with strong immersion, interactivity, and openness.
During the traditional PC internet era, digital human technology mainly focused on CG modeling, inertial motion capture technology, real-person driving, and offline rendering to create two-dimensional flat digital humans. Entering the mobile internet era, technology gradually shifted towards static scan modeling, optical motion capture, real-person + intelligent driving, real-time rendering technology, human dynamic 3D reconstruction, 3D visual motion capture technology, and intelligent driving technology to create three-dimensional solid digital humans. In the future, with the further upgrade of interactive media and the entry into the metaverse era, differential rendering,PBRRendering, light field imaging technology, 4D reconstruction algorithms for motion capture, neural network rendering, and other technologies will become important trends and hotspots in future technological development, leading the evolution of digital human products and technologies.
The development of artificial intelligence technology is a key breakthrough that drives technological innovation in various processes related to digital humans.
Data source: Frost & Sullivan consulting, LeadLeo research institute
The production process of digital humans can be divided into four main modules: modeling, motion capture, driving, and rendering. Currently, it is difficult for the technical solutions of each module to balance precision, cost, and time consumption. The core technologies in the modeling phase mainly include CG modeling, static scan modeling, and dynamic light field reconstruction. At present, there are issues with low precision, high cost, and long time consumption in modeling, as well as conflicts between data collection and lighting presentation effects. The core technologies in the motion capture phase mainly include optical motion capture, inertial motion capture, and visual motion capture. It is difficult to achieve a balance among precision, cost, and efficiency for these three methods, with computer vision-based motion capture being greatly affected by external environmental factors; the driving phase is mainly divided into two schemes: human-driven and AI-driven, with human-driven digital humans being predominant on the market at present. AI-driven solutions have lower costs and time consumption than human-driven ones, but there are still significant deficiencies in lip-sync adaptation and natural language understanding. The rendering phase mainly includes real-time and offline rendering, with current technology struggling to balance timeliness and precision.
Currently, the digital human industry ecosystem is gradually taking shape, with various manufacturers entering the stage of application scenario implementation and business model refinement.
Data source: Frost & Sullivan consulting, LeadLeo research institute
In the digital human industry chain, business models mainly include software and hardware bundled payment and annual subscription. The charging basis primarily depends on the refinement of the digital human avatar and the scope of application deployment. For digital human manufacturers, the outright purchase model offers relatively stable capital inflows, mainly consisting of software and hardware bundling fees; whereas the shared model is mainly based on public cloud, focusing on software sales, which can expand service targets and improve sales efficiency, thereby reducingmarginal costIn the future, digital human producers can further expand their service scope and obtain higher premiums by presetting more refined digital human templates.
The underlying architecture related to digital humans is currently a key investment focus across various industries. The content direction mainly includes live streaming and television, with leading companies gradually enhancing their ability to attract funds.
The underlying architecture is currently a focus of investment in various industries. In the content scenario field, investments mainly enter through live streaming and television. At the investment round level, investments in the digital human sector are generally at the forefront. Among them, there is a trend of later rounds in the film industry, with leading companies gradually enhancing their ability to attract capital, and the market is moving from dispersion to concentration. At the investment method level, establishing subsidiaries or related business departments is a common investment form.
Data source: Frost & Sullivan consulting, LeadLeo research institute
The commercial value of digital humans in the pan-entertainment field is gradually being validated, and they will continue to bring about a revolution in new models for content, social networking, entertainment, and other tracks through the integration of virtual and real worlds.
Data source: Frost & Sullivan consulting, LeadLeo research institute
Digital humans are widely used in the pan-entertainment industry, helping to solve existing problems and pain points in this field. Currently, the products that have been implemented in the pan-entertainment domain with digital humans mainly include digital avatars, virtual idols, virtual anchors, and game characters. Typical application cases include Ling, Guan Xiaofang, Zhiji, and others, who are China's first hyper-realistic virtual KOLs and brand endorsers.
Digital humans have become a key transformation direction for many financial institutions in the future, helping to achieve cost reduction and efficiency improvement.
Data source: Frost & Sullivan consulting, LeadLeo research institute
In the current development of the financial field, there are a series of problems such as low service efficiency, long repetitive processes, low precision in customer demand matching, and limited online time for manual customer service. Currently, the products implemented with digital humans in the financial field mainly include financial assistants, virtual advisors, digital hall managers, etc. Typical application cases include Baidu's financial assistant Xiao Pu, Magic Areva's offline branch digital human Sunshine Xiaozhi, and SenseTime's hyper-realistic AI digital hall manager 'Xiaoning'.
The domestic digital human market has now formed a relatively complete industrial ecosystem map.
The upstream of the digital human industry chain consists of digital tool/technology providers (companies specializing in image design, pre-shaping, production tools, and AI), midstream includes digital human manufacturers (general/computer network technology companies, specialized AI companies, CG companies, and AR companies), and the downstream mainly involves digital human demanders (companies providing enterprise services, entertainment, and various other sectors).
Data source: Frost & Sullivan consulting, LeadLeo research institute
The upstream of the virtual digital human industry chain includes hardware manufacturers such as display devices, optical components, sensors, and chips, as well as software manufacturers like modeling software and rendering engines. The midstream of the industry chain comprises software and hardware systems, production technology service platforms, and AI capability platforms, providing technical capabilities for virtual digital human creation and development. The downstream integrates virtual digital human technology with practical application scenarios, including industries such as media, gaming, film and television, finance, culture and tourism, forming overall industry solutions that empower the development of various fields.
Virtual digital human technology, products, and their integration with different industries are merging with people's production and life. Their scalability, customization, and replicability can drive improvements in traditional processes, enhance efficiency, reduce costs, significantly boost business efficiency, enhance user experience, and bring about changes to traditional fields.
Evaluation of the Comprehensive Competitive Strength of Digital Human Market Manufacturers
Frost & Sullivan assesses the competitiveness of digital human enterprise products at the initial stage, exploration stage, development stage, and maturity stage. Seven typical digital human manufacturers are identified from the industrial ecosystem map, namely: SenseTime, Baidu, Little Ice, Tencent, Huawei, iFlytek, andXiangxin Technology, and conducted a comprehensive competitiveness evaluation of digital human milestones.
Shang TangWith original multimodal interaction, deep learning, and a full-stack approachAIGCThe (AI Generated Content) technology has built three engines: intelligent generation, intelligent driving, and intelligent interaction, creating a 'one-stop' AI digital human production line.
BaiduSince 2019, the company has been deploying digital humans in areas such as finance and broadcasting. Through cooperation with influential enterprises like Pudong Development Bank, China Central Television, and the Meteorological Administration, it has launched digital human avatars for promotion, followed by the development of its own digital human products to achieve commercial implementation.
Xiao BingThe company's predecessor was the Microsoft Xiaoice team. Since its split from Microsoft in July 2020, the Xiaoice team has accelerated the integration of global cutting-edge technologies with localized products, expanding its leading edge in basic artificial intelligence research.
TencentSince the first collaboration on digital humans in 2018, different departments under the company have launched various digital human products in the pan-entertainment field. They have also enhanced their influence by creating digital human avatars for influential enterprises such as the Forbidden City, Bilibili, and Xinhua News Agency. Currently, they have achieved solutions in many fields.
HuaweiIn 2020, digital employees began to be used both internally and externally within the enterprise, with the first virtual digital human making its debut in 2021. Digital Content Production LineMetaStudioHas profound technical accumulation in the field of digital humans.
iFlytekIn 2017, we launched Zhiyix Assistant for the first time. We have always had rich experience and a mature system in the field of virtual anchors. In 2021, we launched the industry's first virtual human interaction platform, providing one-stop digital human solutions for enterprises and individuals.
Xiangxin TechnologySince the launch of its first virtual host in 2018, and the continuous introduction of multiple digital human images such as virtual idols and virtual anchors, AvatarX, the latest intelligent digital human platform launched in 2022, provides various digital human solutions for various industries.
