Shulou(Shulou.com)11/24 Report--

Do you like digital people?

This question seems quite difficult to answer today. If you look at the macro trend, the positive elements of digital people seem to have reached an unprecedented height. For example, some market analysts predict that the market size of China's AI digital people will exceed 10 billion yuan by 2026, and the overall market will show a trend of rapid growth. For example, in July this year, Beijing issued the Beijing Action Plan for promoting the Innovation and Development of the Digital Human Industry (2022-2025) (hereinafter referred to as the "Action Plan"), which means that the digital human industry has ushered in China's first special support policy. It is even mentioned in the Action Plan that the scale of Beijing's digital human industry will exceed 50 billion yuan by 2025.

However, if we look at it from a micro point of view, we will find that the development and application of digital people are not plain sailing. On Singles Day this year, a large number of businesses began to give up the use of digital people; the digital workers and virtual idols of many enterprises and institutions have not been heard since they were released; the voice that the price of digital people solutions is too high and maintenance is too expensive can be heard endlessly; many digital people projects that have invested a lot of publicity resources and are expected to lose traffic in a very short period of time.

If we say that digital man has completed the creation period from technology incubation to education market at a high speed from the end of 2020 to 2021, then at the present stage, digital man happens to be in the transition period from industrial chain construction to large-scale application transformation. At this stage, although the overall market is still bullish. However, a series of problems from the industrial chain to the application scenario and acceptance have also been exposed. At first, the digital human market can not be judged by simple "good" and "bad".

Rapid birth and death, the ultimate pull of digital people, can not help but remind us of a book-Mo Yan's "fatigue of life and death".

Mo Yan said that to live is to be tired. Today, we can say that not only people, digital people want to live, but also very tired.

Born in a live broadcast to discuss the current situation of digital people, we must first reach a consensus: what is digital people?

Generally speaking, digital people refer to virtual characters with anthropomorphic or real appearance, behavior and characteristics. The point here is that today many people default that digital human is an AI-driven virtual image with the technical features of TTS speech synthesis, multi-round dialogue and semantic understanding. However, in the actual context of the development of the whole industry, digital people have little to do with AI technology for a long time, or only use relatively basic AI technology, such as speech synthesis.

The earliest digital people basically appeared in the form of virtual idols. Like the first sound of the future, Luo Tianyi and so on, it is basically realized by animation design and speech synthesis. At this time, most of the digital people are used as a display, lack of interaction, and the threshold is very high, so it is difficult to popularize.

Then, with the development of the live broadcast industry, a large number of digital people realized by human voice, TTS speech synthesis and lip phenotypic prediction began to appear. At this time, the main role of digital people is still virtual anchors, but more of a "new type of performance" similar to live anchors can not replace human anchors in terms of productivity.

With the continuous maturity of NLP and knowledge graph technology, digital people have the ability of multi-round dialogue. At the same time, the lip capture and facial capture technology brought by machine vision is becoming more and more perfect, and the AI digital human driven by AI technology begins to appear. At this time, one of the most profound changes has taken place in digital people: digital people can complete characters independently, and artificial dubbing and real-life acting are no longer necessary. This allows digital people to align with a huge commercial space: AI replaces human beings in jobs that require simple dialogue and interaction.

There are many such jobs, such as AI digital people first appeared in the navigation and navigation work in the enterprise park. Next, counter service, intelligent customer service and other areas began to appear digital people. Especially in finance, insurance and other fields, because the demand of customers is relatively fixed, and digital people are not easy to make human errors. Under the premise of relatively high salaries in these industries, digital customer service has shown certain advantages.

But compared with the customer service scene, what really makes digital people popular is still live broadcast. From virtual idols to virtual anchors, live broadcasting is always the natural hotbed of digital people and the driving force and hope of digital people's "life". As live streaming with goods has become the mainstream form of e-commerce consumption, anchors have gradually become scarce resources. Many manufacturers and brands have chosen the all-weather, multi-line high-intensity live broadcast mode in the past two years. This leads to a surge in demand pressure for anchors. Coupled with the fact that the epidemic has led to increased uncertainty in the work of anchors and increased costs such as anchor potholes and employment fees, enterprises and MCN institutions have begun to explore alternatives for anchors. In addition, in the current live broadcast economy, the live broadcast needs of many brands are only to spread information and launch products without interruption, so it is possible to replace human anchors with AI.

As early as Singles Day in 2021, we can see that the publicity focus of various e-commerce platforms, live streaming platforms, and MCN organizations has changed from live streaming capabilities to virtual anchors, digital people's live delivery capabilities, and comprehensive cost advantages.

On the other hand, digital anchors driven by AI can stimulate users' sense of curiosity and novelty, so there is generally positive feedback in the live broadcast market. These feedback have supported a great deal of policy support and market predictions on digital people today.

But there is a prerequisite for digital people who are born in live broadcasting and become live broadcasters: the application cost of digital people must be lower than that of human anchors.

This is because, in the final analysis, the digital people who appear as substitutes are not as good as human anchors in question and answer ability, adaptability and talent ability. The prosperity of digital people + live streaming is based on the fact that a large number of enterprises and commodities are waiting for broadcasting, but there is no special demand for anchors, which leads to a large number of low-cost development and template change of digital people to take up their posts quickly.

As the demand for anchors begins to adjust, the cost of human anchors is reduced, and the demand for live broadcasting by enterprises is gradually slowing down. The market benefit of digital people will also gradually fail. From another dimension, the development cost and production cost of digital people are rising, and the cost advantage of human anchors begins to loosen.

This is because digital people have encountered the biggest challenge of their career: homogenization.

Die of homogeneity at least at this stage we must admit that AI-driven digital humans are still in the stage of limited interaction and single expression. In the case of awkward dialogue, digital people play more of the recorded content, or switch to the "puppet" mode of human dubbing + speech synthesis + lip recognition.

Of course, the value of AI for digital people is diverse, not limited to multiple rounds of dialogue and knowledge graph, but also visual and voice content. But in any case, the digital human application soon entered a basic dilemma: because there are only a few sources of AI models behind the digital human, even if each digital person has a different design in appearance, his interactive ability and function are mostly the same.

Good-looking skin is one in a million, but boring soul is the same. As the first wave of novel experiences brought by digital people gradually fades away, how to make digital people different has become a great challenge.

As a result, we can see that many digital people disappeared not long after their debut. "the death of digital people" came much faster than expected. The digital people of a clothing brand have not been heard from after only being active on live streaming and short video platforms for a few months; the digital people promotion ambassador of a large event has been complained by netizens because the modeling is too rough; a large number of new virtual idols lack characteristics that can be remembered; many studios summon human anchors back after trying digital people for a period of time.

The vast majority of digital people swarmed up did not stir up any splash, but were submerged in a high degree of homogenization of image, function and interaction, and disappeared in the twinkling of an eye.

On the one hand, it is optimistic about the long-term development of digital people, on the other hand, homogenization is eroding the application space of digital people. If you want to escape this strange circle, the industry seems to have to roll up. To get rid of homogeneous competition, digital people need volume design, volume sound excellence, volume rendering and, more importantly, volume AI technology. The core technologies of digital human include graphic computing, AI motion capture, NLP, speech synthesis, knowledge breakthrough and so on. These technical capabilities have room to upgrade, but the cost is the greatest.

Once rolled up, if you want to get rid of homogenization, the cost of digital people will rise in a straight line, and then lose the low-cost advantage of standard human anchors. This dilemma of prisoners' dilemma has led the digital industry to enter a period of fatigue before it can expand its ambitions.

Digital people, also tired to survive, have to roll, this is the eternal proposition of the science and technology industry. For today's digital people, it is indeed very cheap and negligible for enterprises to quickly get a digital anchorman with a change of skin and shell. However, once enterprises are dissatisfied with such digital human products and want to be able to make creative and intelligent digital people, the cost will increase sharply.

This kind of cost pressure has led to the fatigue of using digital people in enterprises.

Specifically, this kind of fatigue comes from many sources. The cost of supply chain in the early stage, the cost of operation and maintenance in the later stage, and the cost of pleasing end-users all aggravate the difficulty of long-term application of digital human. We can sum up these pressures into three kinds of fatigue of digital people:

1. Industry chain fatigue Digital people have a very long and complex supply chain. Since it needs AI model as the underlying support, it also needs the support of development tools, platform environment and deposit network resources. In addition to technical capabilities, digital people also incur costs in design, animation, professionals and so on. If an enterprise wants to obtain a perfect and high-quality digital human solution, it either has the ability to integrate multi-faceted supply chain with technology development and art design, or directly buys a customized digital human solution.

Obviously, the latter approach is more realistic for most enterprises. But customization and high requirements mean that the upfront cost of digital people is soaring. Whether these costs can be recovered in subsequent applications, or whether they can be effectively reduced, is unknown for enterprises. This is why many enterprises would rather abandon digital people directly than upgrade digital people. Or still let the rudimentary digital human solution struggle.

On the other hand, for digital human supply chain enterprises, a large number of digital people flock to live broadcast, which really brings fire to the business. But the over-focused business has put the next upgrade into a dilemma. From the perspective of industry chain, digital human may be involved in display hardware, optical devices, chips, AI models, AI development tools, modeling software and other technical fields, as well as application layer industry developers and software suppliers. So many complex industrial chains end up serving live streaming only. A single application scenario is difficult to absorb the innovation cost of the industrial chain. This makes it difficult for the underlying technology of digital people to upgrade quickly or dilute the cost.

Too long, too complex, and lead to mobilize the whole body of the digital human industry chain, making it difficult for digital people to make changes with low cost and high efficiency. The industrial chain is afraid that the application window is too narrow, and the application is worried that the upgrading of the industrial chain is too expensive, which finally makes digital people feel anxious and anxious under the good situation.

two。 Another problem of operation and maintenance fatigue in the current digital human industry is that many enterprises find that digital people can afford it but cannot afford it. This is because the overall cost of customized services for digital people is too high, and many suppliers are worried about the market situation. As a result, it is adopted to reduce the cost of an one-time solution, but to bundle a business model supported by operation and maintenance. As the commercial supply and demand of digital people is very simple, most of this model needs to be paid for by e-commerce who rely on virtual anchors.

This taste-and-buy model is very easy to bring anxiety at the later stage of operation and maintenance. With the habit and even boredom of users, the direct value of digital live broadcasting decreases, which will be highlighted because of the cost pressure. For enterprise users, they will increasingly take the continuous operation and maintenance costs of digital people as a burden, and have the idea that they are unwilling to upgrade and renew. As for service providers, they are more willing to reduce the real operating costs of digital people and expand their actual profits. As a result, many digital man painting styles are getting worse and worse, and the experience is becoming more and more unintelligent. This is because the service provider reduces the cost of AI model or image rendering of digital human. What's more, some service providers even send live dubbing instead of AI technology in digital people without informing Party A, with a view to reducing operational pressure. It is also unexpected that AI has not succeeded in pretending to be a real person, and that a real person has begun to play AI.

3. Whether it is the early input or the late cost, the fatigue of the digital human industry comes from the cost pressure. The source of all the cost pressure lies in the aesthetic fatigue of the end-user to the digital person + live / short video mode. After all, the interactive ability and understanding ability of AI are limited, so it is difficult to achieve a steady stream of new value output. However, users' expectations for anchors are unlimited, monotonous and repetitive, which has become the stereotype of digital anchors by many users.

Users with increasingly demanding tastes will eventually promote the reshuffle of the digital human industry chain. The low-cost, high-repetitive model is difficult to maintain. Whether digital people can really enter the stage of large-scale application depends on whether they can jump out of this strange circle of fatigue.

If the dubbing is Digital Man 1.0, the lip shape is Digital Man 2.0, and the AI driver is Digital Man 3.0. Then the digital people who are moving towards the 4.0 era must face to get rid of the stereotyped functions and move towards a broader industrial space beyond live broadcasting. Intelligence and difference are the key to cure the fatigue of digital people.

At present, the idea to solve this problem is gradually taking shape. For example, using large models to increase the "brain capacity" of digital people, the emergence of ChatGPT quickly excited many industries, including digital people. More plans to improve the level of digital human intelligence by pre-training large models will become the focus of upgrading in this field.

For example, through hardware upgrades, to achieve a wider range of commercial value of digital people. The new generation of VR hardware has become very popular in Europe and the United States, and has begun to spread to China. In the VR scenario, digital people can interact with users in more ways, thus bringing new requirements and solving some of the aesthetic fatigue in mobile live streaming, which is also a way for digital people to get rid of the dependence on live streaming.

In addition, continuously shortening the digital human supply chain and reducing the upstream cost of digital human is also one of the solutions. Many cloud computing manufacturers have begun to launch packaged one-stop digital human building solutions for enterprises and software developers, so that digital people can buy out from the solution and transform more low-cost solutions of independent development and self-customization.

The final results of these problem-solving ideas remain to be tested in the future. At least today, there seems to be no reason to think that digital people have ushered in spring. In the face of prosperous market expectations and policy expectations, digital people should be asked a few questions first:

1. What is the value of the digital man in the live broadcast? How long can their irreplaceability and cost advantage last?

two。 Former head anchors are also out of date. How long will it take before users get tired of digital people?

3. Besides broadcasting, is there really any commercial value that can prop up such a complex digital human industry chain?

Digital people really want to go out and need some new technology to bring about a large-scale and long-lasting market for new applications. Perhaps only a deep-seated technological change can cure the fatigue of digital people.

This article comes from the official account of Wechat: brain Polar body (ID:unity007), author: Feng ci Yuan

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.