Wen 丨 Lu Jiu Finance
From the perspective of user experience, economic model, and maturity of the industrial chain, in a very long period of time, virtual people live broadcasting will be a false proposition. However, Alibaba, Baidu, iFLYTEK and other major manufacturers are still actively deploying, racing in all aspects of technology and industrial chain ecology.
Virtual human live broadcasting has not yet become a mature track, but there is no shortage of runners.
On January 14, 2022, Li Shiyan, head of Baidu Intelligent Cloud AI Human-Computer Interaction Lab, said that he hopes that in the past two years, through the opening of Baidu's intelligent Cloud Xiling platform, everyone will realize the freedom of digital people. He mentioned that the important application scenarios of "Virtual Man 3.0" are video, live broadcasting, chatbot, etc.
According to a data as of December 2021, Taobao now has more than 200 virtual people live broadcast rooms, and this number will grow rapidly.
According to a report by Huachuang Securities, iFLYTEK's AI virtual anchor solution has successfully cooperated with a number of media, and in addition, it has a high practicability for online live broadcasting platforms.
So, what is the status quo of the virtual person live broadcast that these big manufacturers attach so much importance to?
Walking into the virtual anchor live broadcast room, at first glance, it is not only both physical and mental, it will introduce goods and goods, but also a dance and Rap to adjust the atmosphere and mood. More importantly, it is very safe, there is no risk of collapse like celebrities and head anchors, it will not get sick, lose temper, get tired, and even broadcast continuously 24 hours a day.
However, Lu Jiu Finance found that virtual people live broadcast, on the one hand, the user experience in the live broadcast room is poor, there is no interaction, and the car often rolls over, on the other hand, for most ordinary businesses, it greatly increases the operating costs, and it is still a failed economic model.
At this stage, virtual anchors can't grab the jobs of real anchors. Baidu, iFLYTEK, Ali and other technology manufacturers, the current vigorous investment, layout is the future.
Virtual anchor = real person fake head?
To form a mature track, you can't just rely on Luo Tianyi, A-SOUL, and Lingling to have a few star virtual people. So, what is the experience of watching the live broadcast of ordinary virtual people in the live broadcast room?
At present, ordinary virtual people do live broadcasting, basically carrying goods. To this end, Xiao Yu, a colleague of Lu Jiu Finance, went to the virtual anchor live broadcast room for a real experience, and he concluded: "There are three main types of virtual live broadcasts, one is to give real people a cargo anchor 'set fake head', one is a virtual anchor with a real assistant to carry goods, and another is a 'robot' virtual anchor with goods similar to the official flagship store of vivo mobile phones." ”
First of all, it is the live anchor with the goods "set fake head". This has become a mature small industry, and the real person behind the virtual anchor has a technical term for this - called "the person in the middle", which is literally "the person inside". Xiao Yu saw in the official flagship store of JuneYao on Douyin that the virtual female anchor vigorously introduced her own products in front of the camera, but during the live broadcast, the virtual person actually sorted out her nails.
Xiao Yu after a long time of viewing experience, so summed up: "The first time I entered the live broadcast room, I feel very novel, because it is generally a real person, the motion capture special effects in its live broadcast room have a strong sense of follow-up, and you can also see the subtle expression changes in the face of the anchor, but if you look at it for a while, you will find that there will be obvious shadows and tears at the junction of its neck, and sometimes the virtual special effects will suddenly disappear, and the sense of violation is still quite strong, I feel that this is not the kind of face change we often play? ”
According to the data on the evening of January 13, 2022 in the JuneYao live broadcast room, there are 6279 fans in the fan group, receiving 5000+ likes, but they have not entered the Douyin belt goods list and the hourly popularity list in the same time period, and there are viewers in the live broadcast room who ask the anchor: "How does your face look so strange?" ”
It can be seen that in such a form, the audience does not seem to buy it.
The virtual IP in the field of food, "I don't eat in vain", takes the form of virtual anchor + real assistant to bring goods, it will interact with the real assistant in the live broadcast room, and introduce a series of gourmet goods. After Xiao Yu experienced it in his vibrato live broadcast room, he felt that this form was more fun, and the two people could have a lot of brilliant points to watch.
According to the data of Douyin, its live broadcast debut achieved a douyin fan increase of 120,000+, a week of 1.26 million+ fans, and in just two months from the premiere of six million to reach the doubling of the size of fans, its goods such as Di Ice Cream has reached the top of the Douyin good things list within 24 hours.
Although the record is very commendable, it has not essentially escaped its dependence on real people.
Xiao Yu also experienced Tmall's vivo mobile phone official flagship store, the live broadcast here uses the form of a full virtual person, the anchor is a cute Q version of the character, in the live broadcast room to explain the performance of their own mobile phone, the voice is gentle and kind, from time to time will also dance for the audience to live the atmosphere. However, when Xiao Yu wanted to interact with the anchor, he found that the virtual anchor was like a robot and could not interact at all.
Xiao Yu asked her, "What is the activity of link 18?" She also continued her dance steps, and after the end, she began to explain the unfinished products just now, and did not communicate with Xiaoyu in language and action, but replied in text on the public screen: "The problem you just mentioned, you can contact our store customer service staff to help you follow up." ”
Xiao Yu's conclusion of this experience is: "This type of virtual anchor feels like intelligent customer service, very mechanized, probably the biggest sense of interaction is just when I enter the live broadcast room to welcome it, and then what to do, with the format interaction she requires to evoke her voice interaction, although it is a supplement to the form of real anchor, but I feel that I can not completely replace the sense of companionship of real people, she said buy it, do not miss and other marketing words, plus mechanical cavity, feel lost soul." ”
Costs increase, and the economic model does not hold
After lujiu finance and a number of industry insiders exchanged, everyone's common view of the virtual anchor with goods is that the economic model based on the current virtual anchor is not established, it has not launched any new solutions compared with the previous live broadcast with goods, did not expand demand, but increased costs.
At this stage, the dependence of virtual anchors on real people is still relatively strong. "I used to hire a streamer in the live broadcast room, and now I have to spend an extra money to buy a virtual human technology." Lion, a senior e-commerce practitioner, told Lu Jiu Finance: "The brand uses the model of a pull-over virtual anchor or a virtual anchor + a real assistant, which is equivalent to our dual employment. Moreover, the form of live streaming is essentially unchanged, and the gimmick is only temporary, which does not really increase sales, and actually does not have much significance. ”
According to the GF Securities Virtual Person Research Report, the cost of creating a virtual person Internet celebrity or anchor is very high in terms of technology and operation, and the upfront investment cost of MCN Company to create a video virtual person is as high as one million, and the production cost of each updated video in the follow-up maintenance process is equally high.
Mr. Wang, a virtual person technology integrator, told Lu Jiu Finance: "The current virtual anchor with goods is still a pseudo-proposition, which greatly increases the cost but cannot increase the income, which may just start to give people a bright feeling, but after the scenery may be a bubble." ”
Moreover, in the current operation mode of the virtual anchor with goods, it is actually a test of the cooperation between the operation team behind it and the "middle man", and if you are not careful, the advantage that the virtual anchor will not collapse will also disappear.
In the previous interactive session of the virtual anchor with goods live broadcast, there were cases of crisis caused by improper impromptu speeches of "middle people", such as the "middle people" of Akai Shin and Kiryu Coco, two virtual idols under Holdlive, who made two consecutive inappropriate remarks, which eventually led to all Japanese virtual idols on holdlive being withdrawn from the Chinese market.
In addition, the technical limitations at this stage are also the main problems restricting virtual anchors. Due to the price threshold of motion capture equipment and the limitations of application scenarios, ordinary end users currently have no incentive to purchase high-end motion capture equipment or customize virtual human images. In the previous 618 promotion, Hatsune Mirai, as a top-level virtual anchor, could not escape the problem of modeling rigidity, and even turned his head directly into a bald baby.
Many Taobao live virtual anchor teams also said that many devices have a battery life of only more than 3 hours, which cannot meet the 5-6 hourS Taobao live broadcast powder increase model, which is not conducive to powder increase and development.
Mr. Wang said: "If the virtual anchor cannot bring new growth value and application in the business model, this matter has no meaning compared with the real person with the goods, more than a commercial gimmick." ”
The nodes of the industrial chain are relatively fragmented
Lu Jiu Finance learned that at this stage, the industrial chain of virtual anchors is probably composed of three links: the upstream is the technology manufacturer that provides the underlying technology, the midstream is the technology integrator responsible for technology integration and integration application, and the downstream is the brand merchant that purchases virtual anchor services.
According to the GF Securities Research Report, the upstream technical parties, including rendering modeling, motion capture surface capture, AI technology, XR equipment manufacturers, etc., are responsible for completing the production of virtual people. The consolidation of technical competition barriers is inseparable from the early large amount of cost investment, internet manufacturers compared to a single layout, small and medium-sized technology manufacturers, it has more advantages in the capital and resource investment of virtual human technology, and at the same time has more ability to create an aggregated technology platform, which can greatly achieve cost reduction and efficiency increase in the production of virtual people, and aggregate a series of precision hardware, technologies, algorithms and software.
Mr. Wang said: "At present, these companies that provide technology I think can not reach the state of the inner volume, because their focus will be different, the threshold of virtual anchor technology is still relatively high, each technical link has a lot of companies in development, the underlying technology should be handed over to large factories to develop, such as the virtual anchor to use the voice interaction, midstream integration can purchase iFLYTEK, Baidu, Microsoft Xiaoice services, etc., mainly to the cost, compatibility and openness of the technology itself as a reference standard." ”
The midstream is the content output link of the virtual person track, which includes asset production companies, planning and operation companies, etc., responsible for the planning and operation of the virtual anchor IP, and providing the upstream technical party with the human design, image design (or demand) of the virtual person. Upstream technology companies use technology such as modeling and motion capture to produce corresponding avatars and deliver midstream content production companies.
"I think we are very much like the role of the shelves in the supermarket, when implementing a plan, we may use different software and hardware of dozens of companies around the world at the same time, select products and services according to the actual needs and budgets of customers, and write new plug-ins and tools on the basis of the underlying technology." Mr. Wang explained.
In addition to providing technical integration services, some of the content parties in the midstream are also responsible for operating the virtual anchor IP, docking the downstream rich application scenarios and partners, and finally realizing the monetization of the virtual anchor. For example, BlueFocus is a typical manufacturer of this, which has jointly created a virtual live broadcast room with Alibaba Damo Academy, and the merchants who buy its products are concentrated in well-known brands such as Unilever, Keep, and Mead Johnson.
According to the "Virtual Digital Person In-depth Industry Report", China's current virtual digital person market attracts players from the upper, middle and lower reaches of the track. However, the efficient synergy of the industrial chain still needs to be tapped and improved.
At Baidu's AI Open Day on January 14, 2022, Li Shiyan publicly pointed out that the biggest problem in the virtual human industry at present is that the industry cannot be coordinated, there is no effective integration between companies, and a complete industrial chain has not been formed. At present, most companies in the industry are only one or several links in the whole process of digital human production and operation, and the industrial ecology has not yet been opened, which makes the cost of making virtual people very high.
Where is the future of virtual anchors?
So, where should the future of virtual anchors go? Or is it that the big factories are competing for in the virtual anchor industry?
First, there is a breakthrough in the underlying technology. The qubit industry report pointed out that virtual digital people are not suitable for areas that are too versatile, professional, and interactive, and will expose their existing technical shortcomings. Huachuang Securities also said in its research report that the current virtual people are highly dependent on scripts, or can only respond to simple random problems, and deep advanced artificial intelligence technology still needs to be developed and mature.
It can be seen that the big factories are constantly making breakthroughs in the underlying technology, Baidu intelligent Cloud Xiling application of four major AI engines, namely portrait drive, natural dialogue, voice interaction, intelligent recommendation level to achieve digital people can hear, speak, understand and interact. iFLYTEK launched its self-developed AI intelligent interactive all-in-one machine, which combines AI core technologies such as speech recognition, semantic understanding, speech synthesis, and avatar drive to achieve "face-to-face" interactive communication between users and virtual characters.
For virtual anchors, NLP (natural language processing, solving the interaction problem between humans and computers) among the three core technologies of artificial intelligence is the underlying technology that needs to be broken through, including knowledge graph, lexical analysis, syntactic analysis, semantic analysis and language model.
Lujiu Finance found that in these five technology fields, Baidu, Tencent, Huawei, Alibaba, and Microsoft frequently appear in the top five positions in the number of patents.
iFLYTEK revealed to Lu Jiu Finance that the live broadcast room is a very direct commercialization scenario for virtual anchors, and whether virtual people can bring greater economic value to the live broadcast room requires the exploration of the entire industry, and iFLYTEK is exploring with the partners of the iFLYTEK open platform.
In addition, you want to expand the application scenario of the virtual anchor. Li Shiyan pointed out that at present, the service scene of digital people and the acting scene have not been effectively connected, and the virtual anchor is not only a sticky scene, but also an efficiency scene.
The third generation of digital people's modeling and content production are all involved in AI, which makes the production of digital people more efficient, more intelligent, and can be oriented to more application scenarios. It can be used for all groups of users, can be carried by video, live broadcast, chatbot and other carriers, and has the ability to interact with thousands of people.
At the end of January, the Communication University of China will launch the "Annual Development Report on the Influence of Virtual Digital People in China", which divides virtual digital people into three categories: virtual idols, virtual anchors, and corporate IP. With the expansion of virtual digital person application scenarios, these classifications will increase, for example, virtual customer service is a large application category, and "real virtual avatar" will also be a massive market.
Finally, it is the layout of the industrial chain ecology. Li Shiyan believes that the effective way to solve this problem is to build an ecology with a large platform and an industrial chain, which will eventually produce a flywheel effect and reduce the cost of virtual people.
Baidu's previously launched intelligent Yunxiling digital person platform integrates digital human survival, content creation, and business configuration services, and can provide one-stop virtual person creation and operation services for radio and television, interactive entertainment, retail and other industries. In addition to Baidu, in terms of building a virtual human ecology, Meta in the United States and Tencent in China are also in the layout.
Li Shiyan said that the ultimate pursuit of virtual people is that you are communicating with him, but you cannot find out that he is a virtual person. Only in this way will the virtual talent be warmer, more humanistic, and the content provided will be more vivid. Maybe that day is the moment when the virtual anchor track officially starts.