Douyin "dislikes" digital people

When digital human technology was first introduced, it was disliked by everyone because of its high cost. However, after several years of development, the cost has been reduced, and its 24-hour online feature has made it another option for live broadcasting. This year, the "Hundred Model War" took place in China, and low-cost digital humans entered the market. Digital avatars at the level of one hundred yuan were quickly hyped up as a profitable business.

At 1 a.m., a female anchor was still guiding users to buy group-buying coupons in her live broadcast room. She was the anchor of Wangli Restaurant, and her longest record was 13 hours of continuous sales. Occasionally, there were questions from passing viewers, but she did not respond.

This anchor is not a real person, but a digital person that Wang Li bought for more than 2,000 yuan, with the right to use it for one year.

"The cost is much lower than hiring bloggers to visit the store." Wang Li was initially very satisfied with this newly hired "employee", but now he has lost all confidence. "Digital Human Company said that it can broadcast 24 hours a day, but the actual situation is that the account is often banned by the platform, and I can't contact the sales for after-sales service."

Wang Li felt that he did not make money or attract traffic, but was instead being ripped off. He was not the only one, as many businesses said that their use of digital people for live streaming was deemed illegal by the platform and their accounts were blocked. Many people, like Wang Li, had no after-sales service.

At the end of last year, with the support of AIGC, virtual digital humans have rapidly become popular since Liu Run, the founder of Runmi Consulting, and other financial bigwigs revealed that they use digital avatars.

This year, the "Hundred Model War" took place in China, and low-cost digital humans entered the market. Digital avatars at the one-hundred-yuan level were quickly hyped up as a profitable business. Every night, when you open the mainstream platforms (TikTok, Taobao, Kuaishou, Video Account, JD.com, Pinduoduo), you will see a large number of digital humans sticking to the live broadcast room.

However, in the past few months, doubts about digital humans have gradually emerged, such as fake effects, chaotic market, opaque prices, etc., and the platform is also strengthening supervision. On May 9, Douyin issued the "Douyin Platform Specifications and Industry Initiatives on AI-Generated Content (hereinafter referred to as the Digital Human Industry Initiative)". So far, other platforms have not followed up.

Douyin's regulations have sparked a lot of discussion in the industry. Zhou Yilu, an operator close to Douyin, told Dingjiao that this means that Douyin has clearly opened up AI-generated pictures, videos, and derived virtual human live broadcasts, which also means that digital humans will be "regulated."

Douyin was the first to take action because, on the one hand, digital people are the most popular on Douyin, and the cost of controlling content risks is very high; on the other hand, digital people’s live broadcasts do not contribute much to Douyin’s business. For example, Douyin’s interest e-commerce focuses on emotions and randomness, and digital people are no match for real people.

Digital people that are well-known to the public include those with IP, such as virtual idol Luo Tianyi and virtual blogger AYAYI, and those with functional functions, such as digital customer service of digital banks. This article discusses digital people that help companies and anchors sell goods and shoot videos.

Different companies may call it differently, including digital human, virtual human, virtual digital human, digital clone and even AI anchor. A large number of digital humans on the market are not connected to the AI ​​brain, so they are uniformly referred to as digital humans in this article.

Digital humans are the easiest field for AIGC technology to apply and monetize. Too many people are eager to use them. Companies want to use them to reduce costs and generate revenue, and anchors want to use them to assist in creation and grab traffic. But many people have the same question as Wang Li: Why are accounts easily blocked? Why is the digital human market so chaotic? Do these platforms welcome digital humans?

1. Blocked digital people are prone to four major pitfalls

"Dude, you've been scammed. This is a recorded broadcast." Wang Li sent the video of the digital human live broadcast to several industry groups for advice, and someone told him this.

As digital humans swept through major live broadcast rooms, a "ban wave" also occurred at the same time. According to many interviewees, recording is one of the most common pitfalls for digital humans in live broadcasts.

Because digital humans for recording and broadcasting are rampant on the market, it has lower costs and relatively backward technology. You only need to write a script, enter the corresponding software, let the "digital human" generate videos according to the script, and play the video during the live broadcast.

"Many of the first batch of businesses that tried digital human live streaming technology were cheated by recorded live streaming." Zhang Xian, head of a digital human agency, said that many businesses had never done live streaming before and were not familiar with the platform rules, but were very interested in digital humans. They spent tens of thousands of yuan to buy a digital human with pre-recorded videos from agents, but they were easily blocked."

In fact, whether it is a real person or a digital person, recorded broadcasts are severely cracked down on Douyin and Taobao Live. Song Xinxin, an operator of the live broadcast platform, said that this is because the platform requires real-time response according to the on-site situation during live broadcasts, but recorded broadcasts cannot do so.

Douyin's "Digital Human Industry Initiative" mentions that digital human live broadcasts must be driven by real people for real-time interaction, and interactions driven entirely by AI are not allowed.

Chen Hong, founder of 51 Digital Human, told Dingjiao, “If the live broadcast content is repeated for a long time, or a set of live broadcast scripts are used repeatedly in a short period of time, it may be judged as a recorded broadcast and banned.”

If there is a "look-alike" face, the platform may also identify it as a recorded broadcast, and the probability of being blocked will also increase.

"If you see several accounts using the same digital human, it means that you have most likely purchased the same digital human model." Zhou Yilu said that this greatly affects the user experience, because in theory it is impossible for the same person to appear in multiple live broadcast rooms at the same time and talk about different content.

Gao Cong, an entrepreneur in the AI ​​field, told Dingjiao that it costs over one million yuan to make a sufficiently realistic and phenomenal digital avatar, such as Liu Run, while most small and medium-sized enterprises can only afford digital humans priced at hundreds or thousands of yuan. Most people choose public models, which are easy to have similar faces and voices.

The current product models of digital humans are divided into public models (public models), anchor customization and source code systems . The prices range from hundreds to tens of thousands of yuan. The main difference is the authenticity of the image and voice.

Fixed focus mapping

Among them, public models are the cheapest. The mainstream digital human platforms on the market all provide a certain number of public models, with monthly fees ranging from tens to hundreds of yuan.

Those who want to use their own image will generally choose the anchor customization mode. The price has experienced a dive, from hundreds of thousands of yuan to the current hundreds to thousands of yuan. The later video generation needs to be charged separately according to the length of time.

The outside world believes that the most expensive is the third model, namely the digital human system source code or SaaS software. After purchasing, you can modify the digital human independently. The quotes from many companies are around 50,000 yuan.

The lack of real-name registration and identification is also the reason why some digital human accounts have been blocked. Douyin mentioned in the "Digital Human Industry Initiative" that virtual humans need to register with real names on the platform and identify virtual human live broadcasts for the public to judge.

Chen Hong told Dingjiao that the registration process is "not easy to go through" for some small and medium-sized enterprises.

However, from the perspective of the platform, the risk control of digital human content is indeed becoming increasingly serious. It is difficult to identify the authenticity of content generated by artificial intelligence, and some digital humans are highly realistic. If black and gray industries take advantage of this loophole, it will be difficult to hold them accountable.

Li Min, a special researcher at the E-commerce Research Center of the China Internet Network Information Center and a senior partner at Shanghai Han Sheng Law Firm, once pointed out to the media that the main purpose of Douyin's release of this specification is to regulate the AI-generated content on the platform and prevent it from causing infringement and other issues. At the same time, Douyin's release of the specification is conducive to reducing the platform's liability after an infringement case occurs. The current law stipulates that if the platform fails to take necessary measures for an infringement case on the platform, it needs to bear joint and several liability with the infringer for the expanded losses.

Zhou Yilu said that some accounts were banned because the content infringed upon copyright, violated regulations, or involved fictitious false advertising and fraud.

In addition, the poor quality of digital human content will also attract the attention of the platform.

Chen Hong summarized the three major manifestations of low-quality digital humans: lip movements do not match , or the audio and picture are not synchronized; the video quality is rough , 720 resolution is used to impersonate 1080 HD, there are green edges on the cutouts, the character movements are unnatural, and the actual delivery and advertising cases are not consistent with the pictures; the interaction is poor, mainly manifested in the use of recorded broadcasts to impersonate live broadcasts , slow Q&A feedback in the live broadcast room, and no real-time interaction, and the use of real human voices to drive the background does not save labor costs.

If a customer's account is blocked, Zhang Xian said he would remind the customer to take the product offline and provide guidance on how to unblock it, "but if the problem cannot be solved, we have no other options." He reminded that no company can guarantee that its account will not be blocked.

2. Digital people are popular, but also chaotic

The reason for the increased supervision is that the digital people in videos and live broadcasts are of varying quality. The good ones are hard to tell whether they are real or fake, and are even used as a tool for profit; the bad ones are obviously fake, affecting the user experience.

Combined with the analysis of many interviewees, because this industry has only been developed for a short time , the technology side is far away from the market, which can easily lead to a mismatch between products and market demand. In addition, digital human technology has the characteristics of " easy to enter the market but difficult to go deep ".

Chen Hong sorted out the recent high-quality digital human cases for "Focus", including Liu Run's digital human avatar, Wanda's digital human anchor, and the digital insurance agent made by Face Changing Technology for Mingya Insurance, etc. Many people sighed, "It turns out that digital humans can be made so real." There are also a large number of digital human cases from nearly 100 digital human companies such as Silicon Intelligence and Fengping Intelligence, which jointly promoted the development of the industry.

The most excited are two groups of people. One group is small and medium-sized entrepreneurs who originally wanted to make short videos but were restricted by the production threshold, or had limited energy and lack of expression ability . What makes them excited is that they can have a digital avatar for a few hundred or a few thousand yuan.

Technology companies that export digital human products are also very excited, but many small companies have incomplete sales channels and lack customer resources; large companies are unwilling to invest too many service personnel because the customer price of 2D digital humans is low and the after-sales costs are too high. There are no specialized agency companies in this new field, so another group of people, that is, agents with a large number of customer resources, have entered the market in a high-profile manner.

The product is made by the Digital Human Technology Company, and the agent is the middleman between the customer and the product. Because the digital human market has great potential, many agent companies provide one-stop services, from selling digital human products to offline digital human training, with the goal of winning over small and medium-sized enterprises that want to do short videos and e-commerce. During this period, many agents used digital human videos for sales and lectures, which was full of gimmicks.

The result of all these factors is that not only are the prices of digital human products opaque, but it is also difficult for the supply side to see the real market demand, and the products produced cannot solve the pain points.

According to Chen Hong's observation of the market, the core demand of customers for digital human products is to pursue authenticity as much as possible within a budget. However, many digital humans on the market "give explanations in a loop for 24 hours, do not drink a sip of water, do not smile, have no atmosphere, no interaction, and have poor authenticity and experience. Even if users see them, they will probably just swipe past them."

Many people think that digital humans are "too fake" because there are too many entrants into this emerging market, but with limited technology, not many high-quality products are produced, and the user experience is poor.

Types of virtual human technology Source/ China Business Industry Research Institute

Nowadays, most digital people in videos and live broadcasts are 2D real people. Compared with the 3D hyper-realistic Liu Yexi and the 2D cartoon A-SOUL, they have lower production costs and shorter production cycles, but the actual effects are mixed.

The broadcasting-type digital people have met the basic experience because of their unified explanation language, and are typically used in the fields of knowledge payment and health knowledge; but the interactive digital people in the live broadcast room are still relatively "fake".

"The mainstream interactive method of digital human live broadcast is to prepare a standard question bank. After triggering the keywords, the answers are directly selected from the question bank and let the digital human 'speak' them out," Chen Hong introduced.

Huang Bo, CEO of Huameng New Media Group and chairman of Taobao Alliance Chengdu, told Dingjiao that instant interaction is the key to transaction conversion in live broadcast rooms. The promotional focus of many companies is that "digital humans can replace real anchors and interact with users", but there are very few digital humans that truly have AI automatic dialogue and automatic chat functions.

"Ding Jiao" asked questions through the comment section in the live broadcast rooms of several leading brands that use digital people. One type of people did not give a response; the other type could give a response like "XX received your question, please wait for a reply", but either gave a reply that was irrelevant after a few minutes, or there was no reply after waiting for five minutes.

Many interviewees mentioned that if the problem of poor interactivity cannot be solved, the market space for digital humans will be limited. Currently, a large number of live broadcast rooms use real people as anchors during prime time and use digital humans as supplements during off-peak hours.

Gao Cong told "Dingjiao" that the next generation of digital human live broadcast products will develop in two directions: improving people and connecting to large models . Both are to improve the interactive experience and be closer to real people, but the current cost is relatively high.

3. Will the digital people on “Tik Tok” become obsolete?

"Small and medium-sized business owners have been ripped off, agency companies have made a lot of money, digital human technology companies have become popular, and a number of star startups have been created." Song Xinxin summarized the current digital human industry chain. As for the subsequent development, in addition to technological progress, it depends on market acceptance, and the main demand parties are watching the attitude of the platform.

Customers use digital humans simply to reduce costs or generate revenue. "In the past period of time, merchants in cross-border e-commerce and local life fields (mainly catering, hotel and travel) have been enthusiastic about digital humans because cross-border e-commerce has a more urgent demand for cost reduction and local life merchants are in the stage of onlineization," Zhou Yilu analyzed.

Even MCNs (hosts and internet celebrities) have set their sights on digital humans. Cheng Cheng, the person in charge of the e-commerce platform, told Dingjiao that some leading MCNs such as Qianxun are making products related to digital humans, not making a digital version of "XXX" for the host, but making it into a business case to sell to merchants.

Agents, MCNs, and the market are all conveying a view that store broadcasting is the main venue for digital human live broadcasting . Many businesses are indeed very interested in digital human products, but their biggest concern, apart from unsatisfactory results, is account suspension.

"As far as I know, many businesses are afraid of being blocked ," Wang Chen, a cross-border e-commerce merchant, told Dingjiao. He spent 30,000 yuan half a year ago to customize a digital human from a digital human technology company. At that time, the other party promised that the platform would not block him, but he was still blocked repeatedly during live broadcasts on TikTok.

He added that the cost of a digital live broadcast room is several thousand yuan, which is indeed lower than the cost of a real person live broadcast, which costs at least 20,000 yuan per month. However, with the risk of being blocked, it is not cost-effective. The most important thing is how much traffic the platform will push to the digital person and how the conversion rate will be, which are all questionable.

Zhang Xian said that after platforms such as Douyin stepped up their supervision, the sales of digital human products have been significantly affected in the past three months; because the cost of trial and error is too high, many customers who are currently in talks are more reluctant to try digital humans with big accounts.

In the past, the platform's attitude towards digital people was ambiguous. Because the platform's KPI is user experience and commercial conversion, digital people are not the optimal solution for the platform to achieve KPI. Song Xinxin analyzed the "focus" and said that the platform hopes that the content pool will continue to be enriched and the creation threshold will be lowered. Ideally, digital human videos and live broadcasts can meet this requirement, but the current situation is that the quality of digital human content is unstable and it just extends the duration.

Recently, the development of this industry has exceeded expectations, with various forces pouring in and even chaos breeding. It is necessary for the platform to increase its control efforts, but if the platform's supervision is too strict, it may also "push" traffic and merchants to other platforms.

Nowadays, the attitudes of mainstream video and live broadcast platforms are becoming increasingly clear: Douyin launched the "Digital Human Industry Initiative"; Taobao opened digital humans to merchants, with prices higher than third parties but a lower probability of being blocked; and Kuaishou and JD.com are both pushing digital humans as a functional tool to influencers and MCNs.

However, due to the different characteristics of the platforms, there are still differences among them . Taobao Live, Kuaishou, Pinduoduo, and JD.com are generally more relaxed; Douyin has the strictest platform rules.

"Both Douyin and Taobao Live are live e-commerce, why do they have different attitudes?" Zhou Yilu analyzed the "focus" and the core is to see how the live broadcast of digital people contributes to the commercial conversion of the platform . Douyin's interest e-commerce focuses on interest, emotion and randomness. The current digital people are completely incomparable to real people; Taobao Live is a shelf e-commerce. Users enter the live broadcast room with a purchase goal. Even if they see that it is a robot selling goods, as long as the price is cheap, it will not affect the order.

A large number of practitioners are trying to figure out the platform's ideas. Cheng Cheng concluded that it is not complicated. From the platform's standpoint, the platform needs digital human-related tools to improve the efficiency of real-life content, but it is by no means a complete replacement for real people . Therefore, if digital people are allowed to broadcast 24 hours a day and try to replace real people, the platform will crack down on it. Otherwise, the platform will allow it.

Digital humans have been previously applied in the fields of film, television, games, and finance. When they enter live broadcasts and videos, digital humans that cost a few hundred yuan each seem to be "flying into ordinary people's homes." But beneath the surface, this market is in a state of explosion and disorder.

Combining the views of many interviewees, like many new things, digital people must go through such a development stage: at first they are sought after and seriously overestimated, then they are questioned after encountering problems, and the platform will rectify them after the problems are concentrated. After adjustments, de-bubbling, and waiting for the technology to mature, digital talents will usher in a real golden period, and only a small number of people will be able to gain a foothold by then.

For a long time before this, the technology side will continue to promote AI and digital people as the general trend. Companies of the same type are very popular in the capital market. Agents always sniff out business opportunities first, and countless merchants rush to enter the market. Platforms continue to find a balance between content and business. There are opportunities and pitfalls in this. We who are in it must always pay attention to and be alert to new changes.

*At the request of the interviewees, Wang Li, Zhou Yilu, Song Xinxin, Zhang Xian, Cheng Cheng, Wang Chen and Gao Cong are pseudonyms in this article.

Author: Xiang Yuan, Editor: Fang Zhanbo

Source: Dingjiaoone (ID: dingjiaoone), deeply influencing innovation.