Working 24 hours a day, are digital people going to take away the jobs of anchors?

Hyper-realistic digital humans are 1:1 digital replicas of real people. With just a computer and a table, you can achieve 24/7 live broadcasting. Will the application of digital humans in the live broadcasting field bring about a new revolution? Let's take a look at the author's analysis.

Chen Feng is a service provider in the wine and travel industry, responsible for the live broadcast business of many large theme parks and cultural and tourism scenic spots. The cost of live broadcasting to attract customers is not low. A live broadcast team needs at least four people: anchor, operation, assistant and field control. If 8 accounts are broadcast at the same time every month, 32 people are needed. Chen Feng calculated that the personnel cost alone is at least 300,000 yuan a month.

After seeing people in the industry discussing digital human live streaming, he bought 10 digital humans for testing. Currently, the highest monthly GMV of a single account has exceeded 10 million. Chen Feng then placed an order for 50 more digital humans for further testing.

The 200-square-meter room is empty, with only AI robots in the computers doing the live broadcasts. "The most important thing is that it saves manpower. The cost of one live broadcast room is only a few thousand yuan, and two operators can be responsible for the normal live broadcasts in 10 rooms," Chen Feng introduced.

The AI ​​robot that Chen Fengsu mentioned is scientifically called "2D hyper-realistic digital human". It is a 1:1 digital replica of a real person. It does not require a venue, only a computer and a table, to achieve 7X24 hours live broadcast.

Not only in the field of live broadcasting, in November 2022, the well-known scholar Liu Run showed in his annual speech that the content on his short video platform was actually completed by "Digital Human Liu Run". Financial writer Wu Xiaobo even said bluntly in his New Year's Eve speech: In 2023, every company may need a digital human, which can be used for local life, customer service, brand live broadcasting, short videos, etc.

Several practitioners told Tech Planet that 2D hyper-realistic digital people are currently used by many small and medium-sized businesses in the local life field. Businesses use digital people to live broadcast and guide users to buy group purchase coupons. With the "May Day" holiday approaching, the demand for digital people in the hotel and travel live broadcast rooms is also increasing.

However, according to most cases presented by practitioners, the GMV amount of a digital person's live broadcast is not too high. For example, in the "Shanghai Disney Two Days and One Night" ticket live broadcast room, the digital person broadcast for 19 hours and the transaction volume was more than 500,000 yuan.

However, through high-intensity, batch account live broadcasts, these digital people can help businesses earn tens of millions of yuan in revenue. A case presented by practitioners showed that a chain restaurant company planned to expand 1,000 digital human live broadcast rooms and open a separate live broadcast room for each store.

Since the ChatGPT concept became popular, people have been paying close attention to which people will lose their jobs and which industries will be impacted. In particular, BlueFocus recently stopped using copywriting outsourcing, which made people feel that the speed at which jobs are impacted by GPT is faster than expected. So, can hyper-realistic digital humans bring about a new revolution?

01 It takes just over ten minutes to create an "avatar" and digital people flood into the live broadcast room

The company that created "Digital Human Liu Run" is Silicon Intelligence. With this digital "avatar", Liu Run can drive the digital human through text, generate a video of himself, and publish it on the short video platform, whether he is at home or on a business trip.

The cost of producing videos using digital humans to generate real-life images is not high. Sima Huapeng, founder of Silicon Intelligence, revealed to Tech Planet that based on video clips of Liu Run's past speeches and through technical training, they can generate a digital human of Liu Run in just over ten minutes, and within one day, this digital human can replace Liu Run to output videos.

The current market price of silicon is 8,000 yuan per digital person. As the technology matures, it will be possible to produce two digital people for 8,000 yuan in the future. "It is a major trend to make digital people that everyone can afford, with low production costs," said Sima Huapeng.

There are many criteria for judging 2D hyper-realistic digital humans, such as facial micro-expressions, synchronization of audio and video, and body movements. Sima Huapeng said that this is like playing hide-and-seek, making it difficult for users to distinguish whether it is a digital human or a real person.

Sima Huapeng presented a set of data: As of now, Siliconbase has produced nearly 1.5 million digital humans, reached cooperation with more than 300,000 merchants, reached cooperation with more than 40,000 e-commerce live broadcast rooms, and has certain cooperation with many big Vs with millions of fans.

At present, according to Sima Huapeng, Silicon Base's annual revenue is expected to reach about 4 billion yuan this year. The annual revenue of the listed MCN agency Yaowang Technology is 4.176 billion yuan, and the annual revenue of the two is already at the same level.

Figure: Digital Human Live Studio, Source: Silicon Base

Digital humans have a wide range of application scenarios, including live streaming of local life and e-commerce, as well as content creation for overseas platforms such as TikTok, Youtube, and Amazon.

In addition to reducing costs and increasing efficiency, some businesses also choose digital humans for reasons such as product attributes. Wang Lei is the head of a live broadcast team at a medical device company. He began to try digital human live broadcasts in November last year. The live broadcast room for medical products involves a lot of medical knowledge popularization, which requires high professional knowledge of the anchor. In the past, Wang Lei's biggest worry was recruiting anchors, and sometimes company employees were "part-time" anchors.

According to the requirements of the Digital Human Service Company, Wang Lei first selected a host with good image and temperament from the company, recorded a 5-minute video, and prepared a speech of about an hour. The host stood in front of the green screen, and needed to speak clearly, have a full mouth shape, fix the excess hair with hairspray, and not make too many body movements. It is best to use 4K resolution equipment and record in a professional green screen background and lighting effect environment. Because the more accurate the acquisition, the better the fidelity and clarity of the digital human.

Then, the digital human service company can use technology to replicate and restore the customer's own facial expressions, voice, and image of a digital human, and synthesize a video of the digital human voice broadcast exclusively for the customer.

Digital people have obvious advantages in live streaming. They are not affected by factors such as time, location, and environment, and can achieve high-intensity live streaming 24 hours a day, 7 days a week. They will not have problems such as getting emotional, getting sick, or leaving their jobs.

Wang Lei said that the company originally formed a live broadcast team of about 20 people. Now he only needs a digital human anchor and a system, and he also handed over the operation work to the digital human service company.

"It's mainly about saving labor costs. The average monthly salary of an anchor is 20,000 yuan, which is 240,000 yuan a year. The customization fee of an AI anchor is 8,000 yuan, and the annual system usage fee is more than 80,000 yuan, which can save more than 100,000 yuan. With 20 anchors, we can save 2 million yuan a year," Wang Lei did some calculations.

At the current stage, digital human live broadcasts usually play an auxiliary role, and many businesses generally do not use digital humans to broadcast during prime time. As a supplement to real-person live broadcasts, businesses will use digital humans to increase off-peak traffic and extend the live broadcast duration, such as choosing digital humans for live broadcasts from 10 pm to 6 am.

Wang Lei's strategy is to use digital people for daily live broadcasts and real people for major festivals and promotions. "Because there is no natural traffic coming in for daily store live broadcasts, it is better to let digital people do the live broadcasts."

Of course, digital human live broadcasts have certain limitations. Wang Lei said that the first is poor interactive capabilities, such as the inability to respond to questions in the comment area in real time and interact with users; the second is that the realism of the image needs to be further improved. "It's okay if you watch it for a short time in the live broadcast room, but if you watch it for a long time, it's still easy to tell that it's not a real person. Moreover, the live broadcast room must have a virtual background, and occasionally the sound and picture will be out of sync. The movements of the digital human are also a few repetitive movements back and forth."

02 Working 24 hours a day, are digital people going to take away the hosts’ jobs?

Liu Zhuyan, the video account anchor known as "Big Cousin", remembers that the Spring Festival movie "The Wandering Earth" made the concept of "digital humans" popular, but starting this spring, the industry began to discuss how to use 2D hyper-realistic digital humans for live broadcasts.

The "Big Cousin" IP has more than 30 live broadcast rooms, covering multiple categories such as beauty, clothing, etc. Currently, 1/3 of the live broadcast rooms are using digital people for live broadcast testing, one is to select low-priced products to ensure sales, and the other is to undertake the sale of some popular products. She also customized a digital person with her own image, which occasionally broadcasts live for her.

"Follow the anchor, those who like more will get benefits later. There are only three sets left, if you don't take pictures now, you will miss them," the upper left corner of the screen shows, with more than 10,000 people watching in real time. However, although the image is realistic, the occasional asynchrony between audio and video still makes users feel that it is a digital person.

According to the current test results, the GMV of a three-hour beauty live broadcast by a digital person exceeds 27,000 yuan. "The most important thing is to reduce labor costs. One operator can be responsible for more than a dozen digital person live broadcast rooms," said Liu Zhuyan.

At present, the price of purchasing digital humans varies from company to company. Sima Huapeng said that if companies do not need to customize digital humans, there are many digital human material libraries available on Silicon Cloud, and the usage fee is 5,500 yuan per month, which includes some AI capability output.

Xie Qiang, the person in charge of the Shanjian Intelligent Market, revealed that if a digital human is used to make a short video for brand promotion, the fee is 298 yuan/month. If it is used for live streaming and uses a corporate model, it is 2,980 yuan/year. If a digital human is customized, an additional customization fee of 8,000 yuan will be added.

The so-called public models refer to the fact that companies often choose to cooperate with model agencies or Internet celebrities and centrally purchase a batch of image licenses as free general digital people.

In addition to software, some digital human service companies represented by Moha also provide agency operation services. Its head Cheng Lei revealed that if it is a local life merchant, about 10% of the transaction amount of the live broadcast room will be taken as commission. In live broadcasting, most of them are individual investors. Moha is responsible for finding the source of goods, creating a digital human live broadcast room and operating it on their behalf, and the final profit is shared equally by both parties.

However, the current interactive ability is still a "hard flaw" of digital humans. Some companies' digital humans have a certain function of replying to public screen questions. When a certain value is reached, some specific actions are triggered, or even the audio reply of the assistant. Some companies will match a two-dimensional assistant in the live broadcast room to cooperate with the anchor and reply to the questions posted by users in the comment area.

Tech Planet noticed that some digital people's live broadcast rooms would pin a purchase method in the comment area, such as a restaurant group purchase voucher introduction, set meals and vouchers can be stacked wirelessly, and the size of a double meal is enough for 3 people, etc. By announcing common problems, the "mental retardation" of digital people can be alleviated.

As digital people become cheaper and cheaper, the unemployment and salary cuts of anchors in Hangzhou have attracted widespread attention recently. As the bonus period of the live streaming industry gradually fades, merchants' investment in live streaming has returned to rationality. An industry insider revealed that in the past, anchors' monthly salary basically started at 20,000 yuan, but it may not be so high in the future.

According to iMedia Research data, the number of registered virtual anchor companies in China increased year by year from 2014 to 2018, and the growth rate basically returned to the pre-epidemic level by 2021, with the number of newly registered companies reaching 948 in 2022. The further development of AI technology will attract more industry participants to enter the virtual anchor market.

Many practitioners said that this type of digital anchor will replace some amateur anchors, but will not be able to replace top anchors like Li Jiaqi, because they have irreplaceable personal IP and dedicated operating companies.

In the opinion of Dabai, the operation manager of the video account anchor "Big Cousin", the irreplaceability of real anchors is reflected in improving fan stickiness and providing emotional value. Digital anchors cannot achieve functions such as livening up the atmosphere of the live broadcast room and enhancing fan participation in the short term.

"Digital human anchors can replace anchors in live broadcast rooms with high content repetition, or anchors in hot-selling live broadcast rooms, that is, live broadcast rooms that only sell one product. Moreover, now the top live broadcast rooms or institutions basically cover hundreds of matrix accounts. Using digital humans to assist real anchors can definitely reduce some costs," Dabai further analyzed.

Xie Qiang, head of the Shanjian Intelligent Marketing Department, believes that hyper-realistic digital human anchors are suitable for small and medium-sized enterprises, as well as mom-and-pop stores and street shops. They have limited budgets and are suitable for purchasing digital human services.

In terms of the style of the live broadcast room, he believes that one type is suitable for "focusing on explanation and not on demonstration", such as snacks and general merchandise; the other type is "users buy for the brand, not the anchor", such as tickets to tourist attractions, Disney tickets, and Mixue Ice City tea. It doesn't matter which anchor is explaining, users mainly buy low-priced group purchase coupons.

According to him, a barbecue restaurant’s live broadcast room can have an average of 30 to 50 people per hour, with nearly a thousand exposures in a day, and a maximum of several hundred exposures per hour. Although it is not extremely popular, it is also an important sales channel for businesses.

03 New trend or false demand?

In fact, live streaming of digital humans is nothing new. The 3D hyper-realistic digital humans represented by Liu Yexi, the hyper-realistic digital humans represented by AYAYI, and the 2D two-dimensional digital humans such as A-soul were all very popular at one time.

Some practitioners believe that the popularity of 2D hyper-realistic digital human live broadcasts is due to the surge in popularity of ChatGPT and AIGC concepts, and the development of local life services on short video platforms such as Douyin. Many businesses need low-cost short video and live broadcast investments.

Taking Liu Yexi as an example, Liang Zikang, CEO of its production team Chuangyi Technology, once told Southern Metropolis Daily that in the more than half a year before the launch of Liu Yexi, the investment in research and development costs, personnel costs, and technical costs was about one million yuan, and the first short video of Liu Yexi cost about several hundred thousand yuan.

As Douyin's local life business represented by takeout and group purchase coupons develops, there are countless small and medium-sized physical enterprises behind it, and they do not have sufficient budget for live streaming.

In comparison, the annual usage cost of 2D hyper-realistic digital humans is basically around 10,000 yuan, which is an affordable way to live broadcast.

Internet giants are also entering the market. At the end of March this year, Tencent released the AI ​​intelligent creation assistant "Tencent Zhiying", which can achieve "image cloning". Users can get their own digital avatars by uploading a small amount of pictures and video materials. In addition, Tencent Zhiying has also connected to digital human live broadcast. Zhiying digital humans can directly replace real people and achieve 7*24 hours of uninterrupted broadcasting.

Liu Wei, the founder of Hualian Digital, said that in the past, the business was mainly concentrated on live streaming of e-commerce products. In the future, the company will focus on the local life field and cooperate with merchants on platforms such as Douyin, Meituan, and Alipay. In his opinion, digital people have great potential in the service field in the future, such as offline shopping guides, offline tour guides, digital doctors, etc.

The differentiated competitiveness of an enterprise depends on the accuracy of the production of 2D hyper-realistic digital humans, Liu Wei said. For example, through a video, can the lip shape, body matching and interactive sensitivity of a real person be perfectly reproduced on a digital human? Whether it is a multi-modal motion generation drive engine, text drive, sound drive or text drive, it can reflect the technological content of the enterprise.

Xie Qiang, head of the Shanjian Intelligent Marketing Department, believes that the industry is still in its infancy and will give rise to professions such as AI digital human operators and AI digital human live broadcast operators in the future.

However, as the live broadcast of digital humans becomes more popular, some mixed bag of companies are also emerging. From the Douyin platform, we can see that companies that use digital humans for live broadcasting are springing up like mushrooms after rain, with slogans such as "new outlet in 2023" and "99% of anchors will be unemployed". Some companies also sell training courses for live broadcast of digital humans.

In addition, are digital humans supported by the platform? Some service providers said that as long as the live broadcast content of AI anchors is good and the conversion rate data is not low, they will not be restricted or banned.

Cheng Lei, the head of Moha, said that his team has set up a special team to target the short video platform blocking algorithm, and will adjust the live broadcast strategy of AI digital people in real time. The probability of "non-real live broadcast" being detected is about 1% to 3%. According to him, at least half of the accounts of other companies will be blocked.

As the person in charge of live broadcast operations for a brand, Xiaoyu will take various measures to avoid the risk of being restricted and banned by the platform to the greatest extent possible. According to her experience, it is best to use video as the background for digital people's live broadcasts. Sometimes during the live broadcast, she will also put recorded real-life videos into the live broadcast room for interaction.

She revealed that some digital human service companies would provide a set of live broadcast equipment, including some technical improvements, to circumvent the platform algorithm, "but this cannot guarantee 100% safety, because the platform algorithm is also upgraded every day."

Bai Chen introduced that there has always been a bad phenomenon in the live broadcast operation industry, which is to test some rules of the platform by reporting the live broadcast rooms of peers. If the live broadcast room is not blocked, it proves that the method can be copied.

In Xiaoyu's opinion, digital human live broadcasts have not yet been widely covered, and many teams are still in the testing stage. Digital humans are just a tool, but the industry has just started, and it has not yet formed an efficient tool. Only as the technology is updated and mature can it be clearly perceived.

"It is true that digital human live streaming has taken off, but whether it can be implemented depends on how it settles," Xiaoyu said.

(Note: Wang Lei, Chen Feng, and Xiao Yu are pseudonyms in this article)

Author: Lin Jing and Yang Xiaohe

Source: WeChat public account: "Tech Planet (ID: tech618)"