Web3.0 Era | How can digital people break through the three major difficulties in scale and landing? Baidu Li Shiyan: Everyone is expected to achieve "digital freedom" within two years

In 2021, which just passed, the concept of the metaverse exploded to the point where it was once "out of the circle" and became the top trend. In the process of global technology giants falling off one after another and the accelerated construction of web 3.0, virtual digital human beings have also attracted much attention as the scene entrance and connection link of the metaverse. At the Jiangsu Satellite TV New Year's Eve party just held, Zhou Shen and Teresa Teng sang "Big Fish" across time and space on the entire network. In the face of the many possibilities presented by digital people, on January 14, 2022, Baidu Technology Salon "AI, I will go! "The fourth issue of the digital human is the theme, showing the panorama of the digital human industry for more than 100 media, experts and scholars.

At this event, Li Shiyan, head of Baidu Intelligent Cloud AI Human-Computer Interaction Laboratory, based on years of experience, systematically shared the current development pattern, technical difficulties and underlying driving forces of the digital human industry. Li Shiyan said that in the virtual digital human industry, infrastructure service providers that provide modeling, rendering, dynamic capture and other services for digital human production have formed a stable pattern. Most of them are controlled by overseas giants, but in the tool and application layer, Chinese companies have Start to shine.

Nonetheless, the large-scale implementation of digital humans still faces three major difficulties: fragmentation of the industrial chain, failure to effectively connect service scenarios and performing arts scenarios, and high costs to meet high-frequency demand. In order to solve these problems systematically, Baidu launched the digital human platform "Xiling". In the future, digital human production, content creation, and business configuration services will be able to achieve full-link production and optimization on one platform, reducing costs and increasing efficiency. "Our goal in the past two years is to let everyone realize the freedom of digital people through the opening of Baidu's intelligent Yunxiling platform." Li Shiyan said.

The large-scale landing of digital people must break through three major problems

At present, the digital human has gone through the 1.0 stage represented by "Paper Man", and the 2.0 era represented by Vtuber anchors has evolved to the 3.0 stage. The digital human 3.0 with high-precision model and artificial intelligence-driven characteristics has become the mainstream of the industry. According to Li Shiyan, AI is involved in the modeling and content production of the third-generation digital human, which makes the production of digital human more efficient and smarter, and can be used for more application scenarios. It can be oriented to all groups of users, and can be carried by video, live broadcast, chatbot and other carriers, and has the ability to interact with thousands of people.

At this stage, Digital Human 3.0 has shown a huge market space. According to the incomplete statistics of Dimensional Spectrum, in 2021, there will be at least 19 financings in the domestic virtual idol/digital human field. In the virtual digital human industry, infrastructure service providers that provide modeling, rendering, dynamic capture and other services for digital human production have formed a stable pattern, and most of them are controlled by overseas giants. However, in the tool and application layer, Chinese enterprises have begun to emerge.

Among Chinese companies, the largest number of companies are asset production, planning and operation companies, such as AYAYI Ranmai Technology, which launched a hyper-realistic digital human, and Chuangyi video, which launched a virtual character Liu Yexi. However, the challenges faced by such companies are that the threshold is relatively high. Low and high substitutability. In addition, Internet companies that master technology and scenarios play the role of comprehensive technology service providers, launching digital human based on their own advantageous scenarios to create application benchmarks; while launching digital human platforms, they continue to implement products in more refined scenarios. , the point and the line advance each other.

Li Shiyan said that the digital human industry has three core driving forces. The first stage of rockets is user demand and technological upgrading, the second stage is policy support and capital influx, and the third stage is the iteration of computing platforms. Essentially, it is driven by technology. Users acquire information and interact from flat graphics, text, audio and video, and the demand for real-time 3D interactive feedback will become the fundamental driving force for the long-term development of digital humans, and will also create a lot of demand in the future.”

Although the digital human industry has a strong driving force, Li Shiyan believes that the current domestic digital human industry still faces three major problems when it comes to large-scale industrialization.

First of all, each node of the digital human industry chain is relatively fragmented and cannot be efficiently coordinated, resulting in high barriers to digital human production and optimization. At present, most companies in the industry are only a part or a few of the entire process of digital human production and operation. ring. Secondly, the service scene and the performing arts scene are not effectively connected, which means that the performing arts digital people do not have the business capabilities required by the customers, and the service digital people lack personality, making it difficult to communicate with users emotionally. The third is that the cost of meeting high-mobility and high-frequency needs is still very high. In the final analysis, this problem is still the production efficiency of digital people.

Baidu launches digital human platform "Xiling"

Facing the current high threshold for digital human production, at the Baidu create conference (Baidu AI Developers Conference) held at the end of December 2021, Baidu CTO Wang Haifeng officially released the "Baidu Smart Yunxiling Digital Human Platform" to build a collection of digital human A platform-level product integrating production, content creation, and business configuration services. At present, the product has the rapid delivery capability of the standard platform, helping customers in different fields to quickly realize agile requirements, reducing costs and increasing efficiency. Baidu's digital human platform Xiling has the characteristics of all-scenario, one-stop, platform-based, and can support from the initial digital human image customization, personality management to business arrangement, content creation, and finally to interact with users in a variety of driving modes It also means that the full-link optimization of the digital human can be realized on one platform, so as to achieve the best effect.

Li Shiyan took the digital human generation as an example to demonstrate the functions of Baidu's intelligent cloud platform. Users can quickly generate a 2D digital portrait that can be driven by AI by uploading a photo on the Baidu Smart Yunxiling platform. The 3D digital human that used to take two or three months to make can now be compressed to the hour level. Digital people can be quickly produced in various forms such as a one-sentence voice description or a simple face pinching. The development of the personalized function of the digital human is also more complete than before. Taking the 3D hyper-realistic oral digital human just released by Baidu as an example, the accuracy rate of mouth synthesis is as high as 98.5%.

In the digital human 3.0 stage spawned by technology, virtual humans can be more closely integrated with different industries. Based on market usage scenarios, digital people can be divided into two types: performing digital people and service-oriented digital people.

Performing arts digital people include virtual anchors, virtual idols, and virtual brand spokespersons who are digital twins. At present, Baidu has created AI sign language anchor, CCTV network virtual host Xiao C, space agency Mars rover digital man Zhu Rong, Baidu virtual spokesperson Xi Gaga, Baidu APP spokesperson Gong Jun digital man.

Service-oriented digital people include digital employees, virtual employees, virtual salesmen, etc., which can help finance, operators, government, retail and other industries to provide a new generation of customer service capabilities based on digital people, solving long queues, slow responses, and high labor costs. problem. At present, Baidu has cooperated with Shanghai Pudong Development Bank, China Everbright Bank and China Unicom to launch a one-to-one service digital human.

Baidu's breakthrough in digital human technology benefits from its continuous layout in the industry. As early as 2019, Baidu started its "virtual digital human strategy". Based on the accumulation of AI technology for many years, AI capability has become the core capability of Baidu's intelligent Yunxiling platform, and now Baidu is also opening this capability to the outside world. Li Shiyan said, "Our goal in the past two years is to let everyone realize the freedom of digital people through the opening of Baidu's intelligent Yunxiling platform."