Sora is popular, are Tik Tok users panicking?

After Sora was released, it directly solved the two major problems in the field of literary videos: accuracy and consistency. In the future, there will be more participants and more intense competition in short videos. In this case, how do short video platforms such as Douyin and Kuaishou respond?

How popular is Sora?

Musk, LeCun, Jia Yangqing, Jim Fan, Xie Saining and other bigwigs are all discussing it, and Zhou Hongyi even said that the birth of Sora means that the realization of AGI (artificial general intelligence) may be shortened from ten years to one or two years.

Many "ordinary people" also smelled the business opportunity in advance, and the business of making money around Sora became popular. According to media reports, a screenshot circulated online showed that a certain institution had sold more than 20,000 Sora training courses and easily earned 2 million yuan.

Today, it has been almost a week since OpenAI released the Vincent video model Sora on February 16, but the discussion is still heated, which is really because it is too excellent.

Functionally, it can generate video content up to 60 seconds long, far exceeding the limit of several to more than ten seconds of other video tools; it can also edit and expand itself, transition two unrelated contents together, and change the style and environment around the protagonist, etc.

In terms of prospects, Sora is not only a video generation tool, it is also seen as a world simulator;

The founding team is also legendary. The core members include 12 people. It has been established for less than a year and is led by two newly graduated doctoral students, some of whom were born in the 2000s.

In addition to excitement, Sora's strength has also made some content creators aware of the crisis. One netizen commented, "Friends in the long and short video and game industries will not be happy for the whole year."

Many industry insiders pointed out that judging from the length of the videos generated by Sora, its arrival will first impact the short video field. Once it is widely promoted, Sora will greatly lower the threshold for short video creation and enrich the content supply of short video platforms, but the problem that follows is that the increase in AI-generated content may affect the content ecology of the platform. At the same time, Sora will also compete with tools such as Jianying and Kuaiying for users, making the previous large investment go to waste.

Sora has dropped a bombshell on the short video field, and the time has come to test creators and TikTok and Kuaishou.

1. Sora lowers the barrier to entry and increases competition

Although there are already many text-generating video tools on the market, such as Pika and Runway, which allow users to input text and get a video content of a few seconds to more than ten seconds, they are not widely used at present. The main reason is that their effects are average.

Dacheng, the head of an MCN agency, told Dingjiao that he had used Runway. "We also purchased high-quality materials from outside, but the content generated in the end was not good enough and could not be used at all."

After seeing Sora, he was very excited. “The length can reach 60 seconds, and the content is logical. It feels like it is on a different level from previous tools.”

"Sora is more than 10 times better than previous similar products," said Niu Xiaobo, a doctoral student in artificial intelligence who has done research on related large models. "Sora's current 60-second stable generation effect will have a revolutionary impact on the short video industry."

Ani, who studies AI, expressed the same view. "The threshold for creators has been lowered. Even if the credibility of the video is not that high and there are some errors, it is difficult to see if you don't look carefully. In the future, technology will be iterated quickly."

According to multiple practitioners, Sora directly solves two major problems in the field of live video: accuracy and consistency.

AI researcher Cyrus feels that compared to previous tools, "Sora is like the Trisolarans fighting the primitive humans."

Accuracy means that the generated video can accurately recognize the text input by the user, and consistency means that during the movement, the main character's content is coherent and does not deform in the multi-view movement changes, and its changes can also follow the laws of physics. The most obvious manifestation is that the interaction between the main character and the background is logical.

A year ago, Niu Xiaobo and his team were still discussing the problem that generative AI could not understand spatial scale, such as the incorrect size of generated images, but they felt that the emergence of Sora had solved this problem.

Some people used videos to restore the physical space model and found that it also corresponded. Niu Xiaobo felt that this proved that Sora's spatial calculation ability was basically correct, and that a large part of people's perception of space also relies on vision.

From this perspective, the videos generated by Sora are supported by the logic of physical space.

Based on the opinions of many interviewees and public statements from many technology leaders, although there are still flaws, Sora's major breakthrough in video and even the entire field of artificial intelligence is that it is a smarter model that even has certain thinking abilities, and the videos it generates can understand the objective laws of the physical world.

This also made a group of practitioners in the short video content industry start to think about whether Sora would take away their jobs.

We simply divide the existing short videos into two categories. One category is made up of real people, who are often the "core assets" of an account or even a company. Sora has little impact on this type of content. The other category is made up of a collection of materials, which tests shooting and creative ability. The emergence of Sora will greatly improve the production efficiency of this type of video.

Some creators who want to increase their output are looking forward to Sora's arrival.

Dacheng said that such tools can completely replace the company's low-end labor. "Our company is small, and we hired a lot of college students to save costs, but they just graduated and couldn't write a copy in half a day. With AI, we can write hundreds of copies in just a few minutes."

Last year, he and his team began using AI tools on a large scale. "Although these tools are still immature, they are very useful for processing copywriting and rough cutting. As a video producer, we must improve efficiency," he said.

For some creative teams, they believe that Sora will not pose a threat to them in the short term.

Dak, head of a leading MCN agency, told Dingjiao, "Sora has not yet started public testing, and the technology is still a long way from maturity. Moreover, our videos mainly rely on people's creativity and brainstorming, including shooting, which also emphasizes innovation. We think about how to design the shots and how to transition. Although Sora is very powerful, it cannot meet our current creative design needs. " He also revealed that the company is not using any AI tools at present.

However, the problem these creators must face is that the emergence of Sora has lowered the threshold for short video creation, and more and more people will join in to compete with them for traffic and monetization opportunities.

A new media practitioner said that they have the ability to create content, but suffer from the lack of a suitable video shooting and editing team. The emergence of new tools is their opportunity.

Based on the opinions of many industry insiders, Sora currently has a relatively small impact on short videos featuring real people, but a relatively large impact on animation and commentary content, and may even eliminate many companies that have no creativity and only take advantage of hot topics. In terms of positions, some people who do mechanical and repetitive work may face elimination, but creative people will be promoted to higher positions.

In short, in the future, there will be more participants in short video production and the competition will be more intense.

2. Will Douyin and Kuaishou panic?

The emergence of Sora will bring more abundant supply to short video platforms, but it will also bring new challenges to Douyin and Kuaishou.

Practitioners said that on the one hand, too many AI-generated videos will have an impact on the original content ecology of short video platforms. On the other hand, Sora's popularity will snatch away some users of tools such as Douyin, Kuaishou's Jianying, and Kuaiying.

Dak told Dingjiao that AI-generated content needs special labeling when uploading to short video platforms.

In May last year, Douyin released the "Douyin Platform Specifications and Industry Initiatives on AI-Generated Content". This initiative proposed that such content should be clearly marked, and virtual human users need to be authenticated with real names. At the same time, strict penalties should be imposed on the use of AI tools to create and publish content that infringes on portrait rights, engages in fraud, etc.

Platforms such as Kuaishou and Xiaohongshu also have relevant restrictions. In February this year, Xiaohongshu officially released the specifications for AI-created content, requiring creators to proactively declare AI-created content and not to use AI-created content to fabricate user experience for commercial realization.

Many practitioners said that they could clearly feel that the platform was controlling such content and even limiting its flow.

"The content created by AI is highly homogenized and not very creative. Once users see too much similar content, they will become impatient and even switch to other platforms," ​​said an Internet industry observer.

He also feels that "if the platform does not restrict this type of content, more and more creators will use AI to save trouble, which will seriously affect the proportion of various types of content on the platform. The most intuitive thing is that some unique and scarce videos will decrease."

And there is also the possibility of a large number of fake videos. Sora can quickly produce a large number of realistic short videos, which makes it easy to fake. Some users find it difficult to distinguish the authenticity of such content. Once deceived, the credibility of the platform will be seriously affected. Last year, many cases of fraud using AI-generated content such as Wenshengtu and AI face-changing were exposed.

In Cyrus's opinion, if the platform supports AI-generated content, it is like "making wedding clothes for others", which will affect user retention. "Imagine that if Xiaohongshu is full of content produced by ByteDance's big model, why don't users go directly to Douyin? And vice versa."

Cyrus predicts that Douyin, Kuaishou, Xiaohongshu, and Bilibili will not recommend a large amount of AI-generated content. Unless it is a large-scale model of a literary video made by the platform itself, it will be monetized separately after being widely promoted by the platform. "For example, when it is optimized to a certain extent, it can directly adopt the user payment model, but none of the products currently have this capability."

More than one industry insider has speculated that the gap between Sora and other AI video creation tools will continue to widen . Previously, the topic of "Will Sora kill Jianying" has sparked heated discussions. If Sora is widely used, most AI tools may be abandoned by the market, and other companies' large initial investments will go down the drain.

Why is the gap widening?

Zhou Hongyi once publicly pointed out that the development level of large models in China is close to GPT-3.5 on the surface, but in fact it is still a year and a half away from 4.0. Niu Xiaobo explained from the technical perspective that Sora is not a point technology, but a complex system engineering. Even though the outside world knows that its core technology is the transformer attention algorithm + diffusion algorithm, etc., behind it is the financial support given by OpenAI and the super data accumulated.

OpenAI has developed a video that can understand the content of pictures through the accumulated GPT general language model and the DALL-E3 image generation model. By the time of Sora, it has further improved its consistency and ability to understand physical space. Each step of accumulation is indispensable.

Bigger predictions make the future of TikTok even more unpredictable. Some film and television practitioners believe that after Sora is widely used, there may be a new platform that is larger than TikTok.

In the view of some people, Sora has lowered the user threshold, which will inevitably lead to a sharp increase in the number of content creators. However, it is difficult to say whether a short video platform like Tik Tok, with its scale, content creation and push logic rules, can carry such content. It is likely that a platform that is more in line with Sora will be born.

Niu Xiaobo agrees with this view after observing the development of the Internet industry. "In the early days, the Internet was used by professional users. Later, the popularity of PCs expanded the user base. Then, smartphones appeared, and everyone could connect to the Internet, so that platforms that serve the needs of all kinds of people emerged. Perhaps as the number of people using Sora increases, a new ecosystem will emerge."

3. Sora has a huge impact, but he doesn’t need to be deified

Currently, Sora has not been open to public beta, and only some visual artists, filmmakers, OpenAI employees and others have been given access. Part of the reason is that Sora is not perfect. Its impact on short videos has not yet been fully demonstrated.

AI trainer Xiaogang pointed out that Sora has the "hallucination problem" that is common among large models , and it still has defects in understanding physical laws and interactions between people and objects.

For example, OpenAI's official website released some defective videos, including a candle that could not be blown out, a basketball exploding during a fall when shooting, and a dug-out chair floating up directly, etc. This is because Sora's understanding is not mature enough, and he mistakenly believes that things dug out from the sand are generally lighter.

When it comes to the short video industry, Xiaogang believes that one of Sora’s most important shortcomings is that it is still unable to fix the characters.

For example, a creator uses Sora to generate a 60-second video with A as the protagonist. However, when generating the next video, the protagonist may be replaced by B.

Compared to Pika and Runway, everyone can see that Sora has strong spatial ability, but no matter how short the content is, it is meaningless if the characters cannot be fixed. Users continue to watch content on short video platforms, whether it is real-life or animated, and a central character is needed to ensure the continuity of the story.

As well as the computing power problem, many people are also worried about the speed of Sora's popularization. After all, no matter how great the technological breakthrough is, if the cost is too high, it will still be difficult for the public to accept it. Many industry insiders believe that this is the biggest difficulty Sora is currently facing.

"At present, OpenAI has not announced Sora's specific computing power, energy consumption requirements and parameter quantity. Preliminary judgment is that the amount of data, computing power and large amounts of energy consumption it requires will be huge costs, which will inevitably conflict with its widespread promotion in the next stage." Niu Xiaobo believes.

Every time a new technology emerges, it triggers a big discussion about the replacement of the old with the new.

When virtual filming technology was first introduced, it saved filming costs for the film and television industry, but traditional filming methods still exist today. More than a year after OpenAI launched ChatGPT, text practitioners have not been replaced, but instead more people have realized the value of high-quality content creators.

Similarly, Sora has brought an impact on similar AI tools on the market, but no matter how powerful it is, it will eliminate more short video practitioners who have no creativity and only want to increase volume, or those who do not know how to use such tools.

As to whether Sora will subvert the entire short video industry and whether the next super platform will be born, everyone is still waiting and watching. It depends on whether Sora can quickly solve the existing problems and whether there will be new challengers in the country.

* The pictures in this article are from Sora's official website. At the request of the interviewee, Dake is a pseudonym.

Author: Wang Lu; Editor: Wei Jia

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