Tech

Can a virtual person at the thousand yuan level replace the anchor?

Top manufacturers are moving toward homogenization

  From 10:00 p.m. to 6:00 a.m. is the idle period recognized by the live broadcast industry. At this time, when you open the live broadcast platforms such as Douyin and Taobao, you can still see the “anchor” with exquisite makeup, who is very patiently selling various delicacies, Group purchases for wine and tourism, but most of them are simply explanations, and there are not many categories involved, except for local life group purchases, which are department stores and fast-moving consumer goods. And how many of the anchors who created these “Sun Never Sets” live broadcast rooms are real people?
  ”Efficient and tireless” digital virtual humans have always been expected to become the strongest “workers” in the future, which is also the original driving force for Internet giants to be keen on “creating humans”. The birth speed of virtual humans is indeed getting faster and faster under the pressure of capital and technology.

  Recently, Tencent Cloud released a smart production platform for small samples of Homo sapiens. According to the official introduction, this platform only needs three minutes of video material to complete the modeling and generate high-definition portraits. The so-called “numerical sapiens”. The most important thing is that the cost of “creating humans” for Tencent Cloud personal version has dropped from millions to thousands of yuan, and the enterprise version that requires high-precision customization has also dropped to about 10,000 yuan, greatly reducing the threshold for using virtual humans.
  However, Tencent Cloud is not the most aggressive player in this field. Beginning in 2021, virtual humans at home and abroad will be born together by the east wind of the metaverse: game manufacturer Epic Games will launch a virtual human creation platform in March 2021, claiming that it can create a virtual human (Metahuman) close to a real person in a few minutes; HUAWEI CLOUD In September 2021, the first virtual human employee “Yunsheng” will be launched, which can carry out simple dialogue and sign language translation; Ali will launch the virtual human Dongdong in January 2022, and enter the live broadcast room to promote the Winter Olympics; formerly known as Microsoft Artificial Intelligence The Xiaoice of the intelligent team has already been “one step ahead” in commercialization, and has launched many virtual people in subdivided industries, such as Sequoia China’s virtual analyst “Hóng”, virtual singer “Luo Tianyi”, and Vanke Group’s digital employee “Cui”. Xiao Pan”, the virtual anchor “Little Cousin” and so on.
  At this stage, most of the virtual people have the defects of “stiff expression, weak dialogue interaction, and obvious roughness of video effect”. Their “personality” is more given by humans rather than created by AI. The significance of publicity is greater than practice. Until AI technology leaps forward to the point where it can produce products at a commercial level on a large scale.
  Since the end of December last year, large-scale model applications in fields such as text and images have been implemented rapidly, causing AIGC products to spring up like mushrooms after a rain. Virtual humans have also begun to accelerate in all walks of life. The competition of leading manufacturers is also “rolling” towards homogeneity. change.
  SenseTime released the “Daily New” large-scale model system before Tencent Cloud, which only needs a 5-minute live video material to generate a virtual human; Baidu also released a digital human based on the Wenxin large model, saying that the company has a large The ability to model low-cost virtual humans; in addition, Xiaoice also announced the technology of large models and small samples of AI digital employee SaaS products.
  In general, in the technical fields of portrait drive, intelligent dialogue and voice interaction, there is almost no gap between domestic leading enterprises. The only difference is the grasp of the subdivided industries, that is, who can create final products that better meet the needs of the industry .
E-commerce live broadcast, the best position for virtual human?

  If a virtual human wants to truly realize commercialization, it needs more application scenarios that are willing to pay. This is the main driving force for virtual human companies to polish the knowledge graph of various subdivided industries, and the live broadcast industry is obviously one of the easiest application scenarios for virtual human .
  Zhang Yi, CEO of iiMedia Consulting, once mentioned that virtual humans can be classified into service-oriented virtual humans and identity-based virtual humans based on the application level. “Service-type virtual people are functional, they can replace real-person services, complete content production and some simple tasks, and reduce the cost of existing service-oriented industries. Identity-type virtual people have identity, and they are mostly presented as virtual IPs or idols. The future virtual world will provide the core interaction medium of human beings.”

  According to a person engaged in the business of digital avatar products, the commercial operation of identity-based avatars is relatively difficult. Even if the production technology can catch up, they still have to face IP operation problems such as human design, content richness, and fan attraction. In contrast, the landing of service-oriented virtual humans is much faster, and the use of virtual humans to replace scenes that consume a lot of time and do not require much creativity is the “lower fruit” in this technological competition.
  He believes that the cost of virtual human small-sample training technology is approaching the critical point for civilian use. In the future, based on the PaaS (Platform as a Service) interface service launched by cloud vendors, more merchants and well-known IPs will choose virtual human products. “Eventually it is very likely to replace more than 70% of the current anchors.”
  But at this stage, the shortcomings of virtual anchors are also very prominent. Although it can reduce labor costs and improve efficiency by working around the clock, the ability of virtual anchors to solve specific problems is really limited, and the process is relatively rigid.
  When you open the live broadcast platform, you can find that the most suitable virtual anchors are either the above-mentioned live broadcast rooms of standard products such as snacks and department stores that “emphasize explanation and light display”, or “users come for brand group buying discounts” with a foundation of brand trust In live broadcast rooms, such as new tea brands, chain fast food restaurants, etc., it doesn’t matter whether it is explained by real people or AI. The main purpose of users is to buy low-priced group purchase coupons.

  The virtual anchor is more like a background board or a simple drainage tool in practical application. It is almost always a few words of product introduction and discounts. Consumers want to know more about coupon verification, logistics and after-sales issues. You can wait until the working hours of the manual customer service or the real anchor goes online to get an answer.
  Can AI support various expressions of virtual anchors? Can it be as smart as a real person, and interact with the audience in the live broadcast room? These are the bottlenecks in the promotion of virtual anchors at this stage. Although ChatGPT has opened up market imagination, avatar manufacturers rely on multi-modal pre-trained large models to support small samples to quickly customize avatar product strategies. The primary goal is to reduce costs and increase efficiency, not to produce a traditional high-precision avatar. The effect still needs to be observed on the ground.
Technology has not kicked the critical point of demand

  ”Repetitive labor will be replaced by AI” is the consensus under this wave of AI. Amateur anchors at the waist and tail have already felt the chill, but can virtual humans create a brand new head anchor? At this stage, it seems unlikely. If you want to create a virtual anchor similar to “Li Jiaqi”, what you need is a professional operation and technical maintenance team. While bringing the goods, you must create a corresponding “personal design” to provide fans with emotional value. The cost is no more than that of a real head There are few anchors.
  At present, there are few top virtual anchors who can compare with the top real anchors in terms of the scale of live broadcast rewards and the number of fans. The virtual idol Luo Tianyi, who spent 30 million yuan in three years, only has a pit fee of 30 million yuan. It is on par with most mid-waist anchors.
  A virtual human service provider once revealed that behind the once-popular virtual anchors on platforms such as Douyin, there are not AI-driven interactive platforms but real people. “Users still like to interact with real people, and it’s good to tell some jokes, AI is too boring.” .
  On the other hand, the attitude of live broadcast platforms towards virtual anchors is not positive. Short video platforms such as Douyin must pursue commercialization goals and balance user experience. At this stage, virtual anchors are not good at interacting, and rigid programmatic live broadcasts can attract some idle traffic. It is not realistic to pursue user retention data. Douyin official Naturally, the recommended mechanism will not give too much traffic.
  ”Virtual anchors are easily judged as packaged recordings and unmanned live broadcasts. During our testing, more than half of the accounts will be blocked because of non-real anchors. The platform algorithm is being upgraded every day, and we can only Find a way to circumvent it, such as inserting a real-life anchor video as the background.” The service provider said.
  Another hidden danger worth noting is that virtual humans are still in the stage of extensive development, and are destined to face legal and ethical challenges in the future, such as the risk of portrait rights management and control and the risk of false information brought about by deep synthesis technology. Virtual humans, still in their savage growth period, are not yet ready to completely replace an industry.

error: Content is protected !!