Douyin's biggest rival has emerged
Tik Tok's opponent appears
Recently, the hottest thing in the Ai circle is probably Sora.
So, what exactly is Sora? In fact, it is a large Wensheng video model launched by the OpenAI team in the United States last week, which is OpenAI's first foray into the field of AI video generation.
As soon as it was launched, Sora instantly became a global hit and became the focus of everyone's attention.
According to the company, Sora can create realistic and imaginative scenes based on text commands, generating high-definition videos of up to one minute in a variety of styles and formats. That is, when the user describes the required video content and gives Sora some prompts, it can generate 1 minute of Ai video content, and it is very realistic.
For example, give it a description like this:
A stylish lady wears a black leather jacket, a long red dress and black boots, and holds a black handbag as she confidently and casually walks down a street in Tokyo with warm lights, neon lights, and dynamic city signs. She wears sunglasses and red lipstick. The streets are damp and reflective, and the colorful lights seem to create a mirror effect on the ground. Many pedestrians walk the streets.
Sora-generated footage
It can then generate video content that can be converted from panoramic, medium and close-up. If it weren't for telling you that this was generated by Ai, you would definitely not have found out. In addition, it allows you to stitch two videos together in a natural way.
What's more, Sora is not a simple video generation model, but a "world simulator". In other words, Sora can understand the laws of the physical world in the same way as a human, for example, knowing that a tank has a huge impact force that can produce the result of "crashing a car", and knowing that when biting into food, the action of "biting" can produce tooth marks and produce a corresponding "bite mark" image.
This comprehension enables Sora to generate more realistic, physics-based video content.
Comparison of "Sora" and another video Ai generation tool "Runway" Video source: twitter@ @Poonam Soni
However, there are some cases where Sora is technically immature. For example, Sora may have difficulty accurately simulating the physics of complex scenes, may not understand cause and effect, may obfuscate the spatial details of prompts, and may have difficulty accurately describing events over time, such as following a particular camera trajectory.
The emergence of Sora is a new revolution in the field of Ai. OpenAI officially said: "Sora is the foundation of models that can understand and simulate the real world, and I believe this feature will be an important milestone in the realization of AGI." ”
The launch of Sora also sent Open AI's valuation soaring, from $29 billion at the beginning of last year to $86 billion at the end of the year. According to people familiar with the matter, OpenAI is discussing a new round of funding, with a valuation of up to $100 billion.
For short video platforms such as Douyin, the emergence of Sora is undoubtedly like a sledgehammer, which may be smashed down at any time. As AI-generated video tools such as Sora become more popular, Douyin short video platforms may face the risk of user diversion and the challenge of their business model.
Douyin's Ai layout
For example, more users may choose to use these AI video generation tools to create and share video content, resulting in some users streaming from traditional short video platforms to AI-generated video platforms, affecting the user base and activity of short video platforms.
In addition, traditional short video platforms mainly achieve commercial operation through advertising and paid content, while AI-generated video tools can produce video content efficiently, with high quality and low cost.
In contrast, content generation on traditional short video platforms mainly relies on manual creation, and the efficiency and quality are difficult to match with Sora, which undoubtedly brings huge competitive pressure to short video platforms.
However, Song Jian, the former chairman of Baidu's product committee, said: Sora is indeed very good, but it is still too early to think that no one will shoot videos in three years, and Douyin Tiktok will be quickly subverted.
From Sora's side, although it has great potential for development, it is still in its infancy, and it needs to strengthen cooperation with other platforms and use external resources to develop and grow.
Soon after its launch, Sora quietly opened an account on TikTok and updated more than 20 newly generated video content. Currently, the account has more than 110,000 followers.
The reason why Tik Tok is settled is because Tik Tok has a large user base and active social interaction, which provides a platform for Sora to showcase. Moreover, the video content generated by Sora is about 1 minute, which is in line with the tonality of the TikTok short video platform and attracts users' attention faster.
It is worth mentioning that in the AI video generation track, Douyin has not always been in a passive state. Four months ago, Douyin launched a new AI tool "Jichuang", which can generate unlimited short videos, product graphics and texts, live broadcast scripts, etc. for free.
Create a page
"Instant Creation" provides 3 major functions and 6 small projects, of which 6 small projects include intelligent filming, AI video scripts, AI live broadcast backgrounds, AI live broadcast scripts, etc., which can basically meet the needs of creators to produce various videos and graphics on Douyin.
Before OpenAI released Sora, Douyin quietly launched a video model product, Boximator, which can accurately control the movements of people or objects in videos through text.
Douyin related people responded that Boximator is currently unable to land as a perfect product, and there is still a big gap between it and the leading foreign video generation model in terms of picture quality, fidelity, video duration, etc.
For Douyin, the layout of AI will take more time. Douyin began to set up a large model team in early February last year, and is still continuing to accelerate the pace of AI layout and application.
According to Tech Planet, several departments under Douyin are increasing investment in the research and development of AI products, including Douyin e-commerce, giant engine and other business units. At present, a number of AI products have started internal testing, such as the Flow department, which was established in November last year, has launched a number of conversational AI products, including AI products such as Bean Bao (Cici) and Button (Coze), and will also launch the AI character interactive APP "Talker", as well as another AI product "PicPic".
The "Stove" interface in the test is subject to the final launch of the product Source: Tech Planet
According to Jiemian News, citing a person close to Douyin, ByteDance is currently recruiting elite soldiers from across the company to support the AI business. Jun Zhu, the former head of product at TikTok and now the vice president of product and strategy at ByteDance, has become the head of product for Flow, and some of TikTok's core product personnel have moved to Flow.
Recently, Zhang Nan, CEO of Douyin Group, announced his resignation as CEO of the group, saying that he would focus on the development of Jianying in the future.
According to people close to Jianying, in the past year, Zhang Nan has tilted most of his energy to Jianying-related business, and personally led the team to seek a breakthrough in AI-assisted creation, and will soon launch an AI product for pictures and videos.
The AI track is changing too fast
Since the emergence of ChatGPT at the end of 2022, many Internet giants have poured into this track and launched their own large models based on their own advantages. Baidu launched Wenyan Yixin, Alibaba launched Tongyi Qianwen, and Huawei, Tencent, 360, SenseTime, JD.com, etc. have all increased investment in AI research and development.
With the in-depth development of large model technology, the AI large model has evolved from the initial iterative process of automatically generating text, automatically generating images, and automatically generating videos.
The advent of the large model has caused a boom in the Chinese market, and now after the release of Sora, many Sora concept stock sectors such as Chinese Online, Yidiantianxia, Guomai Culture, and Insai Group have also continued to rise. It can be seen that the market is optimistic about the prospects of the Sora shares.
But in terms of Ai Wensheng video, it will take more time for domestic giants to achieve a technological leap. Debang Securities Research Report pointed out that computing power support and large-scale high-quality data training are still the basic disks of large video models.
Wensheng video has higher requirements for computing power, but there are bottlenecks in domestic computing power and data, which may limit the development and application of Wensheng video. Although some progress has been made in algorithms and models in China, there is still a gap compared with the international leading level.
At the same time, the unsmooth flow of data elements on domestic platforms will also hinder the development of AI video. A member of the Information and Communication Economy Expert Committee of the Ministry of Industry and Information Technology said that in the future, the circulation of data elements should be promoted as much as possible, and a balance should be found between the prosperity of the data element market and data security.
During the Spring Festival in the Year of the Dragon, Zhou Hongyi, the founder of 360, once said, "When you see the gap, you know how to catch up." The advent of Sora still gave us a bucket of sober cold water. ”
Author | Zhou Wenjun