Source: AI Whale Society
There are approximately 10 million "soulless humans" in the world, with lifelike appearances, rich expressions and movements, and authentic voices in conversations. They are virtual digital humans.
From the first anime-style 3D virtual human Luo Tianyi 10 years ago, the first realistic style virtual human Ada 4 years ago, the popular ultra-realistic virtual idol Liu Yexi 2 years ago, to a series of AI anchors half a year ago, and the virtual human Jing 2 and a half months ago, the evolution speed of the virtual digital human race track is like riding a rocket.
Today, virtual digital humans have entered various industries and taken on various roles. From digital front desks, to digital HR, live streaming hosts, and brand IPs, apart from the roles of founders and CEOs, almost all other roles have been "invaded" by digital humans.
A common view in the industry is that virtual digital humans are the first race track in the AIGC field to achieve scale and have a high ceiling. Data shows that the global virtual anchor market in 2022 is about 107 billion yuan, and the number of digital human live streaming rooms in 2023 has increased by 400% compared to last year. This is only in the live streaming market, and there is even greater market space in the enterprise-level market such as brand endorsements and digital employees.
MetaAGI CEO Wang Linlin believes that in the future, every company will need virtual humans as a digital asset that always belongs to the brand. Companies can save millions in celebrity endorsement fees and even become celebrity endorsement killers.
Currently, there are already two or three unicorn-level digital human companies in China, while in the AI drawing and AI writing fields, there are few startup companies that have developed to this stage.
But is the virtual digital human race track really a golden road? Wang Linlin once operated a virtual human selected for the Tsinghua Metaverse Pioneer Program. This virtual human IP can generate a monthly income of up to 50,000 yuan, but she admitted that even big virtual IPs with millions of fans in the industry find it difficult to balance income and expenses.
On May 9th, Douyin released the "Platform Standards and Industry Initiative for AI-generated Content (referred to as the Digital Human Industry Initiative)," indicating that the largest traffic platform has also begun to regulate this new species of virtual digital humans.
Different challenges are coming one after another, and the industry of virtual digital humans is still advancing in twists and turns. The Chinese-style representative virtual human IP "Tian Yu" has gained over 6 million fans in half a year, becoming a representative of Chinese culture going global; within 2 years, the virtual digital human from World Excellence has already obtained 1,000 large and medium-sized clients, inspiring the industry's development.
The future development of the virtual digital human industry, how to meet challenges and opportunities, and how to solve commercialization landing are all worth paying attention to. AI Whale Society and Meta AGI jointly hosted the "Virtual Humans, the Geometry of Money" event, where Meta AGI CEO Wang Linlin, Element Realm Technology COO Chen Meihuan, World Excellence Technology Marketing General Manager Gu Xinfen, and OneWu Digital Technology Founder Cao Yang shared the latest industry insights on virtual digital humans.
01. The Challenges of Virtual Digital Humans
"Every day, hundreds of accounts are banned, and it's a constant struggle to make money," complained a CEO of a digital human company in an industry group, which led them to move their main digital human operations to the overseas market in 2022. Another well-established digital human company's GPT clone program is also at a standstill both domestically and internationally.
Apart from policy issues, the biggest obstacle is that platforms are not welcoming. As the platform with the most digital humans, Douyin fired the first shot against "anti-intelligence" in May this year by issuing an AI identification order. This is because in the past one or two years, virtual humans have rapidly proliferated, with an incomplete count showing that there are over 280,000 virtual human companies nationwide, resulting in a large amount of low-quality content.
However, in the eyes of the industry, this may be a good thing, indicating that the development of the digital human industry is traceable. "In the past, virtual humans crowded into live streaming rooms, which actually resulted in a large amount of low-quality content."
Chen Meihuan, COO of Element Realm Technology, believes that this is related to the fact that virtual anchors are "too dull." They lack expressions, have no understanding of product promotion, cannot understand bullet comments, and cannot respond to user demands. Lacking "IQ" and "EQ," they are completely different from real human anchors, and the market is flooded with "bad money," leading users to start rejecting these low-intelligence virtual humans. However, from the new rules of Douyin, it is clear that the platform welcomes high-intelligence, high-quality virtual humans.
Since 2019, Element Realm Technology has been laying out AIGC, attempting to solve the above problems through AI production and driving virtual humans. In early September this year, the first high-precision real-time interactive AI virtual human "Jiang Lingfeng" officially appeared at the Service Trade Fair. The full set of self-developed AI systems gives virtual humans "soul" and multimodal perception and interaction capabilities. Through RGB cameras, it can recognize the user's expressions and body movements, and with real-time backend large model computation, analyze the user's emotions and speech, and then generate corresponding interactive actions and text-to-speech in real time. Judging from the crowded feedback at the Service Trade Fair, digital humans with this level of multimodal interaction capability have broken free from people's stereotypical impressions of virtual humans.
From the industry's perspective, this is a typical capability of virtual digital humans 3.0. Virtual humans 1.0 relied on humans, 2.0 were ultra-realistic but lacked intelligence, or had intelligence but could not be ultra-realistic, while 3.0 are ultra-realistic and intelligent. Only based on ultra-realistic digital humans can they avoid being disliked by platforms/audiences and have the ability to replace real human anchors.
In the view of Gu Xinfen, Marketing General Manager of World Excellence Technology, regardless of how advanced digital human technology is, the biggest problem for many companies lies in post-operation.
In her experience, in 2022, many enterprise CMOs/CEOs were very interested in virtual digital humans and attracted a large number of clients for consultation. 2D digital humans cost tens of thousands, while 3D virtual humans cost hundreds of thousands. Many enterprise CEOs/CMOs tried them out and approved the cooperation.
In 2022, Gu Xinfen frequently went on business trips and negotiated many new partnerships. However, after a year of development, she realized that the biggest problem in this industry is long-term operation: "Some enterprise clients or brand users, the digital humans they produced, due to the inability to keep up with post-operation, resulted in their effectiveness not meeting expectations, and their costs may be far greater than the revenue they eventually receive." In Gu Xinfen's view, many digital humans have even become loss-making projects for companies.
Cao Yang, Founder of OneWu Digital Technology, also believes that digital human companies are not just technology companies now, but also need the ability to output operation plans. In the end, they need to give clients a cooperation that can be clearly calculated for ROI, rather than just doing business.
To truly make money with digital humans, one must solve the problem of their lack of intelligence and increase efficiency. World Excellence's solution is AIGC + World Excellence Small Model, allowing digital humans to have a certain level of intelligence and interact with users autonomously. This way, Douyin will not judge the virtual human's live streaming as approaching a recorded broadcast and will not limit the live streaming room.
On the other hand, they have developed a dual anchor system of real human + digital human, where the digital human attracts attention and the real human assists in live streaming. "The existence of digital humans is not just for show, they also need the ability to promote." This ability is based on the accumulation of tens of thousands of hours of live streaming data, the construction of vector databases through cleaning up scripts, and the building of the digital human's expressive ability.
By live streaming simultaneously on multiple platforms, a single digital human can host multiple live streams, allowing enterprises to replace real human anchors and increase efficiency. This is an evolution that has quietly occurred in the industry over the past two to three years, and it also means that virtual digital humans 3.0 have just become a new thing welcomed by platforms and enterprises.
02. AI Transformation Reduces Two-Thirds of the Threshold
As a pioneer in the virtual IP race track, Wang Linlin is one of the earlier adopters in the industry to use top resources such as 3D motion capture equipment, professional CG teams, and collaborations across the entertainment and drama fields to create virtual IPs.
When talking about the virtual human IP she is responsible for, Wang Linlin's team consists of six people, including modelers and operations. The main income comes from Douyin's monthly subsidy of 10,000 yuan and the virtual human's tips. During each live stream, there are usually about 3,000 people online, and after the Chinese New Year, the traffic dropped to 1,000-2,000 people. "We didn't buy traffic, it's all natural traffic." Many viewers think the income is good, but in reality, it's already not easy to break even in some months.
The core issue is still the high threshold. The cost of ultra-realistic digital humans is currently very high. A leading ultra-realistic digital human company in Shanghai has quoted prices up to 700,000 yuan, while the virtual human Wang Linlin is in charge of is self-produced by the team, which also leads to high monthly expenses for the team. "For the Peking Opera version of the virtual IP project, the development of the voice, appearance, and stage took about 4 weeks, and this cost cannot be reduced."
A similar situation is seen with the flagship "Tian Yu" virtual IP from Element Realm Technology, with nearly 300 million video views across the internet, showing outstanding performance in fields such as Chinese culture, beauty, aesthetics, fashion, culture, and art, but correspondingly, the high-level and diversified content production costs will be relatively high.
The evolution trend from CG digital humans to AI digital humans is a big trend in the industry. Only through platformized intelligent production of virtual digital humans can the cost be reduced. For this purpose, Element Realm Technology has developed the Element Enjoy SaaS platform and the Element Fun AI platform, combining AI capabilities to provide users with low-cost creative capabilities. The former allows for the free combination of eyes, mouth shapes, etc., to synthesize digital humans, and facial expressions are also supported by large model algorithms. The latter supports the generation of images, text-to-video, and text-to-speech, AI-izing the entire content production process. "This means that ordinary people without technical backgrounds can also produce their own digital humans."
For example, from the simplest art perspective, designing a character's image or a scene in the past might have taken about 10 days, but now, by combining with AI, this time can be shortened to 2 days, and several versions of such an effect can be produced to confirm with the client, greatly reducing the cost of early communication.
In the modeling stage, using the self-developed "MetaSurfing Element Enjoy Intelligent Cloud Platform," traditional modeling steps are simplified, and the platform's built-in digital assets, such as AI expressions/actions, shorten the production time of virtual humans. Subsequent interactive actions are generated and driven by AI. From actual experience, AI has already revolutionized the production mode.
Gu Xinfen also mentioned that the company has spent the most effort in the past two years on lowering the threshold for digital humans. Currently, World Excellence can use a 10-minute high-definition video to capture the expressions and body movements of real people, and through technologies such as lip editing for talk head, voice cloning, and large language model (LLM), achieve a 1:1 replica of a digital human model. Then, in the system backend, they can select the digital human, live streaming background, and plugins for live streaming room decoration.
According to Gu Xinfen, the total cost of traditional live streaming rooms with real people, assistants, and operations personnel, as well as the purchase of live streaming and filming equipment, and the decoration of live streaming rooms/video shooting venues, amounts to 700,000 to 1 million yuan per year. In the early days, if you wanted to achieve the effect of virtual humans, the cost was also between 700,000 and 300,000 yuan. However, after iterations, the current digital human system has largely achieved AI transformation, and the image and voice of the digital human can be used for life, totaling less than 80,000 yuan per year, reducing costs by more than two-thirds.
03. More Possibilities for Monetizing Virtual Digital Humans
"Scenes, content, technology, products, and commercialization have finally come full circle. It's really not easy, those who understand, understand," recently, Chen Yan, the founder of the virtual digital human company Next World Culture, expressed this sentiment on her WeChat moments. This was a week before the Next World Culture co-branded event with Naixue's Fantexi Milk Tea, which caused the server to crash, and Chen Yan's reflection was made.
As a race track that virtual digital humans have been exploring relatively early in the AIGC field, the actual commercialization is still actively being explored.
The two main commercialization paths for digital virtual humans at present are: the first is the To C model, where the company establishes its own IP and then operates this IP on the company's platform, expanding the possibilities for this IP, such as live streaming, advertising endorsements, and paid short dramas.
Liu Yexi from Chuangyi Technology, AYAYI from Next World Culture, and Tian Yu from Element Realm Technology are all of this type. The production and operation costs of such ultra-realistic virtual IPs are relatively high, and it is reported that Liu Yexi's operation team has reached a scale of 150 people. Many virtual IPs have an operation team of only 3-4 people, and commercialization relies on live streaming.
Currently, virtual IPs have traffic and cash subsidies on Douyin and Bilibili. "For example, if you want to create content advocated by the platform, you will receive a cash subsidy of 10,000 yuan per month, and we also participate in activities on the PK side, such as participating in a live streaming ranking activity, and so on," Wang Linlin explained the commercialization paths for most virtual IPs.
Large IPs rely on the company's resources for operation, such as Element Realm Technology creating the virtual IP "Benshan" and entering the beauty and makeup race track. Benshan is positioned as a top student in the field of chemistry, so she has a deep understanding of the ingredients of beauty products. She once achieved a fan increase of 229,000 from a single video. More importantly, she does not speak carelessly, avoiding the risk of the IP collapsing.
Cao Yang mentioned that live e-commerce on Xiaohongshu (Little Red Book) may be the traffic entrance for digital humans. AYAYI rose to fame in the text and image content on Xiaohongshu, so digital humans should also be able to rise in live streaming rooms on Xiaohongshu, as Xiaohongshu has a strong atmosphere for product recommendations.
Many technology companies, if they are stronger in the productization stage, will choose the second path, which is to provide To B services, helping enterprises establish virtual digital humans. Then, enterprises can use virtual digital humans for roles such as live streaming e-commerce and digital employees.
World Excellence Technology mentioned that in various fields they have worked in, the effect is relatively good in local life. "Whether it's a real person telling me about this product, selling vouchers, or selling experience cards, or a digital human introducing it to me, for me, there is not much difference. As long as I want this thing, I will definitely buy it. Therefore, the virtual humans in the local life race track will show more prominent and obvious cost reduction and efficiency improvement, and their effect is relatively good."
According to reports, a merchant used World Excellence Technology to organize a "Gulangyu Island Day Tour" event, and a live stream could achieve a conversion amount of 40,000 yuan, with an 11% product conversion rate. Making it affordable for small and medium-sized businesses to use is actually a comprehensive test for virtual digital human companies. According to industry insiders, most digital humans on the Douyin platform are provided by companies such as World Excellence Technology and Silicon Intelligence.
Element Realm Technology tends to penetrate and expand into the To C market from the To B market. Previously, Element Realm Technology organized new product launch events for brands such as AMD, Peak, and Shanghai Auntie. The immersive XR launch event for Shanghai Auntie was completed by Element Realm Technology's virtual anchor Xiao Anni, and AMD's virtual exhibition hall was presented using its XR devices.
Whether it's World Excellence or Element Realm Technology, they actually come from the AR/VR and metaverse era, so they can provide different solutions to enterprises. The digital human solution may be at the level of tens of thousands of yuan, while the entire metaverse launch event can elevate the cooperation to the level of millions.
On the other hand, each company is trying to get digital humans out of short video live streaming platforms. Choices include digital magic boxes, LCD screens, and XR devices. Digital magic boxes can be placed in front desks, exhibition halls, and other places, becoming a product for showcasing the company's image. The combination of software and hardware undoubtedly opens up new commercialization possibilities for virtual digital humans.
"In fact, I think that in the To B business, for us, the entire company's To B business has shrunk significantly in the first half of the year," Cao Yang, the founder of OneWu Digital Technology, mentioned. This is also the low point of the development of virtual digital humans, but after the industry crosses the singularity, it will usher in a period of rapid development and maturity, and virtual digital humans will truly resonate with the business.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
