Image source: @VisualChina
Text|Market Value List, Author|Wu Zhanguo, Editor|Jia Lele
In 1903, horticulture first invented the word "cloning", and 100 years later, not only plants could be cloned, sheep, monkeys could be cloned, and even people could be cloned.
However, due to ethical issues, research on human cloning has been restricted. Recently, with the popularity of AI, AI clones have appeared. The AI clone launched by Microsoft Xiaoice is like a real netizen, "he" has his own circle of friends, can accompany you text chat, voice chat, and even video chat.
Further on, AI Sun Yanzi covered "Hair Like Snow", which exploded on the Internet overnight, and Sun Yanzi herself said that her fans had changed the court.
In fact, as early as 1927, the silent movie "Metropolis" appeared in the first AI virtual character in film history - robot Maria, and later Mickey Mouse, Japanese anime, people used hand-drawn animation to show two-dimensional virtual images.
However, at that time, because the technology was not yet mature, the application of virtual humans was still in its infancy, and later with the emergence of computer vision and motion capture technology, virtual gradually entered the exploration stage, and now with the advent of the AI era, virtual humans began to really enter our lives.
Whether it is a simple product consultation, or the live broadcast industry, film and television industry, virtual humans seem to be replacing humans. So, what changes will AI bring to the virtual human industry? What are the opportunities and risks of the virtual human industry? Are virtual humans a good business?
01 AI technology reshapes virtual humans
On July 12, 2022, Luo Tianyi, a virtual idol who has been popular for 10 years, spoke out through AI for the first time: "Sing the song in your heart, I am the virtual singer Luo Tianyi." ”
This marks the dawn of an era in which AI is reinventing virtual idols.
In 2000, the movie Final Fantasy: Deep in the Soul used computer vision (CG) and motion capture technology for the first time to design three-dimensional virtual figures.
At the end of 2016, it was known as the world's first virtual anchor, attracting 400,000 fans in 4 months, and the virtual anchor who opened the "A.I.Channel" channel on Youtube is still a virtual idol driven by real people.
It was not until 2018 that Xinhua News Agency and Sogou launched the world's first 3D version of AI synthesis anchor "Xin Xiaowei", that the change of AI to virtual humans began to enter an explosive period.
The changes of AI to virtual humans include three aspects: appearance, behavior, and thought interaction.
First, AI's change in the appearance of virtual humans is that traditional 3D modeling requires people to design an image through software, while AI generation is a 3D image generated by algorithms and previous data.
For example, Xinhua News Agency's "New Xiaowei" is based on its employee Zhao Wanwei, which first scans Zhao Wanwei's body parts 360 degrees through hundreds of cameras, and captures and records his various forms of expressions and movements. Then the various parts and expressions of the "new micro" were modularized, and reassembled like a Lego.
Finally, the virtual human model built by driving and rendering is realized through the algorithm, so as to achieve a high degree of compatibility between expressions, lip movements, body movements and language expression.
Second, in terms of behavioral expression, AI replaces the traditional human dubbing mode through speech synthesis technology, and AI directly generates speech, including singing, news broadcasting, speech, reading aloud and other application scenarios.
The explosion of AI Sun Yanzi is inseparable from the open source AI speech synthesis model VITS released by the Korean Academy of Sciences in 2021, and the so-vits-svc project developed on the basis of the voice model has AI Sun Yanzi through the training of Sun Yanzi's songs, interviews and other voices.
Third, in terms of thought interaction, the emergence of large models allows AI to think and answer closer to humans through the training, thinking and answering of a large number of human datasets, making the interaction of virtual humans more human-like.
For example, Unreal Engine, a software that makes games, basically realizes the leap of virtual humans from "very similar" to "very real". In April, a game based on ChatGPT, Generative Agents: An Interactive Simulation of Human Behavior, was launched.
NetEase's open-world martial arts mobile game "Against the Water and Cold" allows intelligent NPCs and players to freely generate dialogue, and based on the dialogue content, the characters in the game independently give behavioral feedback.
Then there is Microsoft's Xiaoice, an AI clone developed based on the Xiaoice framework, the first batch of Internet celebrities Hanzo Forest settled in, different user chat content is different, reply content is different, more like real human interaction.
02 Who can make money with virtual humans?
Not only NetEase, Microsoft Xiaoice, there are more players in the virtual human world.
On June 6, the 2023 Apple Developers Conference was held in the Apple Spaceship Headquarters Building, and the mixed reality (MR) helmet released by Apple at the conference once again brought people's vision to the metaverse. With the advancement of technologies such as AI, the metaverse world is getting closer and closer to us.
Virtual humans are the constituent elements and interactive carriers of the meta-universe, as a mapping of real humans, virtual humans use motion capture, AI and other technologies to copy the image of real people (digital twins), so as to achieve stronger interaction between virtual and real in the meta-universe, with a better sense of reality.
At present, the virtual human industry chain can be divided into three layers, namely the basic layer, the platform layer and the application layer.
The base layer provides basic hardware and software support, such as display devices VR/AR glasses, modeling software, and rendering engines.
Most of the basic layer is done well by foreign companies, such as Unity engine, which can not only do virtual humans, but also support animation, games, etc., and many popular games are developed through the Unity engine.
The platform layer mainly includes technical support enterprises such as AI vendors and comprehensive Internet vendors.
Platform layer enterprises are the application of software and hardware technology in the basic layer, and combined with virtual digital human technologies such as AI and computer vision (CG), to provide further technical support and solutions for the application layer.
For example, the four AI tigers such as SenseTime, Yuncong, YITU and Megvii, as well as iFLYTEK, which has done relatively well in the field of speech recognition. Others include companies like Microsoft and Tencent that provide one-stop solutions.
The development company of the virtual Hanzo Forest, Xiaoice, is the platform provider and the provider of the application layer.
The application layer of enterprises is to integrate virtual digital human technology with traditional industries to realize the transformation of traditional industries. Such as virtual host, virtual customer service, etc.
From the perspective of commercialization and application scenarios, virtual humans can be divided into three categories: content/IP type, functional service type, and virtual avatar type.
Content/IP-type virtual humans, such as Liu Yexi, a virtual beauty expert who exploded in 2021, are mainly used in scenarios such as film and television, entertainment, and marketing, and functional service virtual humans are mainly used in industry service scenarios, such as e-commerce virtual customer service, these two types of virtual digital humans are more for the B-side.
In addition to the application of the B-side, the virtual avatar type digital human is also oriented to the C-side, such as the famous consulting expert Liu Run, who has customized his own virtual image, and the avatar can replace him for live broadcast and other work.
There are two types of opportunities for monetization, those who use digital human images to pan for gold and those who sell shovels to gold diggers.
Liu Run's digital avatar belongs to the first type, which allows the digital avatar to undertake the work and create income; Lehua Entertainment's virtual idol girl group A-SOUL also belongs to the first type, and virtual idol girl groups, like real idol girl groups, can also make money by selling tickets and peripherals to ordinary fans.
The Chuangyi technology behind Liu Yexi not only uses Liu Yexi's live broadcast and video to obtain revenue, but also is a shovel seller. Liu Yexi has a technology-driven middle office of 150 people and a strong content strategy behind it.
Chuangyi Technology, with its creativity in rain control, has also created explosive materials that can be used by Douyin's huge number of other users, and as of June 2021, it ranks in the top ten of Douyin's special effects use list.
Now, virtual digital people are becoming more and more intelligent, visual, and more common, so is virtual digital people a good business?
03 Are virtual digital humans a good business?
The application of virtual humans is still in the early stages of development, and the rapid development of the future is almost a certain direction.
According to iResearch's research report, the market size and core market size of China's virtual human-driven industry will be 186.61 billion yuan and 12.08 billion yuan respectively in 2022, and are expected to reach 640.27 billion yuan and 48.06 billion yuan respectively in 2025, showing a strong growth trend.
At present, many companies have indeed gained considerable benefits because of the application of virtual humans.
In November 2020, after A-SOUL became popular, Lehua Entertainment's pan-entertainment business revenue increased from 21.1 million yuan in 2020 to 37.9 million yuan in 2021, mainly due to the revenue generated by the commercial development of the virtual artist group A-SOUL, which also promoted the gross profit margin of Lehua's pan-entertainment business to turn significantly positive.
Xiaoice the AI clone launched by the company, there are two payment models: one is the "emotional mode" is 72 yuan / year, which can interact with the clone voice call and friend circle; The other is the "super mode", 360 yuan / year, the clone will become the user's office partner, providing services such as copywriting.
Taking "Hanzo Forest" as an example, if 10,000 users pay, if all choose the emotional mode, it is equivalent to the annual income of the "Hanzo Forest" clone 720,000 yuan; If you choose the super model, you can earn 3.6 million yuan for one year.
On May 2 this year, foreign Internet celebrity Caryn Marjorie cooperated with start-up FV (Forever Voice) to launch Caryn AI, which is the "virtual girlfriend" of fans.
Caryn estimates that the AI version of herself will increase her income by 60 times to $60 million (about 416 million yuan) a year.
The gold mine is big, but the money is not necessarily easy to fish.
First of all, what we want to discuss here is not some large enterprises at the basic layer, for them, their own technological breakthroughs are strategies, not measured by whether they can have a better monetization model and monetization space in the short term, but more for some companies at the platform layer and the application of virtual humans.
First, the cost is relatively high, but the benefits are not certain.
The virtual people needed on the B side are generally used to replace people's work, such as virtual anchors, which are originally intended to extend the live broadcast time and reduce labor costs.
However, as far as the current AI technology is concerned, it is not yet the same as a real person, and many people engaged in the live broadcast business say that the traffic in the digital live broadcast room is not good. An article in "Leopard Change" said that the staff of Mei ONE (Li Jiaqi Company) said that digital humans will affect live broadcast interaction, which is the biggest advantage of live broadcasting.
If you want to make digital people closer to real people, you need to spend more money, such as Liu Run's digital avatar cost at the million yuan level.
High-quality digital people also need high operating costs, just like the creative team mentioned above that requires more than 100 people, from the explosion in 2021 to now, Liu Yexi's works on Douyin are only 41, and the production cost of the first short video is as high as hundreds of thousands of yuan.
In addition to the tool type, there is also a virtual person of the emotional companion type, and the virtual Hanzo Forest in super mode belongs to a combination of the two.
Whether it's Hanzo Forest or Caryn, they were already well-known people before the launch of virtual humans, which is not easy to imitate. Another risk for this category of digital humans is the loss of user freshness.
People are loyal to a relationship, either because of interests, or because they are like-minded, or because they are afraid of being condemned, or they have no choice. And when there are enough virtual people based on well-known characters, and people can reasonably and legally have unlimited emotional resources, loyalty is naturally forgotten.
Tsinghua University also said in the "Metaverse Development Research Report 3.0" that due to the high cost of virtual humans, if the later operation is weak, there will be problems such as heat decline and loss of user freshness, and the value of IP will fall.
Second, there are regulatory and legal risks.
For example, infringement issues. Experts have proposed that AI Sun Yanzi may be involved in infringement of portrait rights, sound rights, copyrights, and data rights.
For example, the issue of fraud. Recently, the police in Baotou, Inner Mongolia, reported a case in which fraudsters used AI face changing and onomatopoeia technology to pretend to be acquaintances to commit fraud, and Mr. Guo, the legal representative of a company in Fuzhou City, was defrauded of 4.3 million yuan in 10 minutes.
With the development of technology, the virtual digital human more like a human may cost less, more suitable legal and regulatory rules will be introduced, and the virtual human industry will develop more standardized, but at present, the cost and benefit are not equivalent.
Bibliography:
- [1] "2022 Virtual Human Application Scenarios and Practice Report", Jiazi Lightyear Think Tank;
- [2] "Media: Virtual Digital Human + AI, Industrial Acceleration", Huatai Securities;
- [3] "A-SOUL Team Receives Exclusive Interview for the First Time: There Are No NG Opportunities in This Industry", 36Kr Interview;
- [4] "Red Star Exclusive Dialogue with Chuangyi Technology Xie Duosheng: Revealing the Birth and Future of Liu Yexi", Red Star Video;
- [5] "Digital Human "Black Hole": Spend 1 Million to Create People, Spend 8 Million to Operate", Jiazi Lightyear;
- [6] "Agent Digital People Earn Millions a Year? The real business opportunity is still "cutting leeks"", leopard change;
- [7] "Special Study on Virtual Digital Humans: Long-term and Short-term Outlook, IP and Empowerment", Essence Securities