#baiducom
**Baidu will launch a live broadcast platform for digital people, and make digital people more human within one to two years**
*The virtual digital human track is becoming a new outlet for major Internet companies to catch up.*
On July 19, Li Shiyan, head of Baidu’s digital human and robot business, was interviewed by many media including Jiemian News, and introduced the latest progress of Baidu’s intelligent cloud platform and digital live human business in detail.
The platform was officially released at the Baidu AI Developers Conference at the end of 2021. It integrates digital human production, content creation, and business configuration services. It mainly provides virtual hosts for industries such as broadcasting, interactive entertainment, finance, government affairs, operators, and retail. , creation and operation services of virtual employees, virtual idols, and brand spokespersons.
According to Li Shiyan, Xiling currently has four relatively mature sub-platforms: digital human sign language platform, digital star operation platform, digital human live broadcast platform and dialogue configuration platform related to interactive capabilities. Through them, they can support solutions such as radio and television, mutual entertainment, MCN, artist brokerage companies and brand owners, and support the commercialization of the platform.
Baidu believes that the biggest pain point in the digital human industry is that the chain is very long: in addition to modeling, binding, and dynamics, a software company needs to help with integration; if you need voice, you need to find a voice company, and you need to look for visuals. Be a visual AI company, and finally let engineers do the integration.
In Li Shiyan’s view, Baidu is the only company in China that has both visual capabilities, speech and semantics, including computer Turing’s automatic generation of full-link AI capabilities. The underlying full-stack AI capabilities are Xiling’s advantage; Up there are various types of portrait production lines, as well as a person management platform. After the human design is produced, Baidu will meet the needs of customers through interactive services or content production.
This also involves the classification of digital human beings. Baidu divides its digital human products into two categories: service type and performing arts type.
“In our opinion, the first principle of a digital human is two things: the first is interaction, and the second is content.” Li Shiyan explained that interaction is to help customers achieve their goals through interactive means such as question-and-answer, and content is mainly produced through production Short videos, pictures, live broadcasts, etc.
These goals include reducing the cost and increasing efficiency of live streaming during idle time, as well as extending new boundaries on the corporate marketing track.
Therefore, Baidu’s digital human business mainly focuses on three tracks: live broadcast and delivery scenarios, corporate marketing (mainly for the conversion and retention of new customers, etc.), and some things in the direction of entertainment anchors.
Among them, the fastest landing is the live broadcast scene. Baidu said that the digital human live broadcast production platform, which will be launched during the Baidu World Congress in 2022, can realize 24-hour pure AI live broadcast. Digital human can switch makeup, scenes and styles at will. A large number of small and medium-sized businesses create their own live broadcasts through the platform. Digital anchor.
Although live streaming has become an important form of marketing for merchants, the cost is not low, and it needs to bear the costs of venue rental, employer broadcast and the entire operation team.
Baidu has done a survey. Hiring a good anchor in a first-tier city generally has a monthly salary of about 10,000 yuan, and the venue cost is 30,000-40,000 yuan a year. Wait, at least 150,000 yuan is required every year, which is a relatively large burden for small and medium-sized brands.
“With our live broadcast platform, a software can solve the problem, and the cost has dropped by 30% or even more than 50%.” Li Shiyan said.
However, to be a digital human live broadcast platform, there are still many technical difficulties to overcome. For example, in the portrait dimension, Baidu has iterated three versions, with the help of hyper-realistic digital human SaaS software, super-intelligent question-and-answer dialogue system, as well as lip synthesis technology, face binding technology and action system, including the accuracy rate of lip synthesis. 98.5%.
In terms of crucial interactive capabilities, Baidu’s integration of voice, semantics, and vision capabilities into a product requires not only breakthroughs in underlying technologies, but also very strong engineering capabilities.
“We believe that through continuous efforts, there is a very good opportunity to make the expressiveness and interaction ability of digital people approach the level of real people without limit within 1-2 years.” Li Shiyan said.
Official information shows that Baidu Smart Yunxiling already has dozens of customers, including the real-time broadcast sign language anchor of CCTV headquarters of this year’s Winter Olympics, the Mars rover digital Zhurong number in cooperation with the National Space Administration, and the first virtual cultural blog in China. The propaganda official “Wen Yaoyao”, etc., are all based on this platform for design, development, integration and application.
https://finance.sina.com.cn/tech/it/2022-07-20/doc-imizmscv2801789.shtml