"如何缓解字节AI焦虑？剪映 vs 豆包，哪个超级应用更适合你？"

Zhang Yiming put the two most important figures of Douyin in the core position related to AI.

In February of this year, OpenAI released a video generation model Sora. Nine days later, Zhang Nan stepped down from the position of Douyin CEO and focused on Jianying. It seems to be a sacrifice. In fact, just like when Chen Lin was responsible for education business, Zhang Nan was conquering new heights for ByteDance.

An insider of Jianying told Xinmei daybreak that actually Zhang Nan has been in charge of Jianying all the time. She has made some adjustments at the organizational structure level, including streamlining the organization and executive transfers.

About three months later, Zhu Jun appeared at the Douyin modelling conference. His trajectory is similar to Zhang Nan’s. Both established a company acquired by ByteDance and both are key figures in Douyin and TikTok rising. Now, they both invest in Big models. Zhu Jun is the product director of Flow, ByteDance's main AI department.

Actually, earlier this year in April, ByteDance also promoted an important figure Zhu Wenjia. He served as the CEO of Today's headline and now reports directly to CEO Liang Rubo, and is based in Singapore all year round. The AI team he led in is called "Seed" inside ByteDance, and in the organizational structure exposed by social platforms, he is also the business director of Flow.

It’s obvious. ByteDance is repositioning the core executives who "have won hard fights" to AI, with heavy firepower.

Zhang Yiming probably foresaw the shockwave of this productivity revolution very early.

The detection of AI-generated video has already begun. As the largest short-video platform in the world, ByteDance is the most aware of the crisis. Liang Rubo alerted everyone in the internal letter that "AI technology will have a great impact on content creation and may even produce new creation platforms."

ByteDance may not be the fastest respondent, but they seem to be the one who bets the most.

ByteDance needs its own Sora. Maybe this is a new product from scratch, such as the newly launched "Dreamina"; or it may be a quick response to trends, offering ideal solutions like Jianying.

However, if the new AI-generated video creation platform is not from ByteDance... this will be troublesome, and a huge trouble.

"Sora kills Jianying" is an eyeball-catching statement in the traffic era, which may reveal some people's expectations. Those big model companies rubbing their hands are waiting for this kind of video publishing entrance, quietly waiting for the opportunity to strike in a moment.

Deep binding of Douyin and Jianying

To understand the carrier of ByteDance's AI transformation, we need to clarify the relationship between Jianying and Douyin first.

Domestic short-video platforms have their own video editing tools. Douyin corresponds to Jianying, Kuaishou launched Kuaiying, WeChat has Miao Jian, and Bilibili is Bi Jian.

The basic functions of these editing tools are similar. If it is for operating an account, these low-threshold, fool-style tools are enough. Professional players may need PR or Final cut when editing large productions.

Why do all platforms push their own editing tools? There is a motivation to reduce the threshold for video creation, contribute ideas, increase the "submission rate" of ordinary users and, most importantly, have a strong association with the operations mechanism of each platform.

For Jianying, its meaning and value also lie in servicing the content ecology of Douyin more.

LatePost quoted an insider of ByteDance who once said that without Jianying, the works in Douyin would be more skewed towards the PGC (content produced by professional institutions). Jianying improved the overall quality of the videos created by users and made the UGC (content produced by users) more consistent and beautiful.

Jianying has a natural advantage in the scene of AIGC, especially the special effects function, which is deeply embedded in the content ecology of Douyin.

The "special effects" videos of Douyin that have spread around recently are actually a kind of AI creation. The "Similar Effects" operation produced similar effects, greatly reducing the threshold for ordinary users to create videos, and the masses imitated each other, with a large number of similar videos flooding Douyin, creating a hot effect.

Wang Guohui, the former deputy director of the special effects team at Douyin, explained that within ByteDance, special effects were called "props" for a long time. This kind of creation tool is an important tool for Douyin and TikTok, helping ordinary people to overcome psychological barriers and express themselves freely, and special effects also have stronger following and demonstration effects, which Douyin emphasizes.

At the same time, ByteDance has strong team support for refined operations. At some important festivals and critical nodes involving user growth needs, the product and operation team will participate in the design of special effect functions in advance to attract user participation and creativity, increase the submission rate, allow more people to be seen and stimulated, participate in follow the shooting, and form a positive cycle.

But what if this AI "special effects" does not come from Jianying, but from other creation platforms? And what if these things happen again and again?

The consequence is that Douyin may lose its video creation entrance. In the long run, it may even lose its publishing entrance.

ByteDance is too familiar with this path. When TikTok started abroad, it was seen by many creators as a good and easy-to-use video editing software. The Instagram reds who were active on TikTok at the beginning were first attracted by the dazzling and fool-style "special effects" and edited them before uploading them to Ins or Youtube. Later, they gradually stayed on TikTok.

ByteDance can do it, theoretically, everyone else can do it too. The big model is surging and Sora came out of nowhere, increasing this variable greatly.

An insider of Jianying told Xinmei daybreak that "the company definitely values it from a strategic level, not limited to AI single-button video creation, but also includes AI material processing, AI efficient editing and other functions." His view is that, after all, it still has a long way to go from AI-generated video to actual landings, just as GPT hasn't fundamentally changed the employment status quo of text workers.

CEO Tan Dai of Huoshan Engine said in an interview with the media that he has no lineup to achieve overtaking on bends for ByteDance's layout in writing and generating videos fields, "There is no shortcut for model training. If the foundation is not done well, it will be a 'AI idiot'—because it cannot even understand text, how can video be good?"

Zhang Nan's New Mission

Zhang Qiao is the person in charge of a media video business. One day, they were preparing to release an analytical video on a hot news topic, but the colleague in charge of dubbing had an emergency and couldn't respond in time.

She noticed that Jianying had previously launched some AI functions and tried to find a solution. There is indeed a "cloning voice" function. As long as a live person inputs a few sentences, AI can copy and output it. She tried it out. If the speed is adjusted fast enough, the effect seems to be good. The breath is not as natural as a real person, but it is better than expected.

But after sharing the finished film with colleagues, most people responded that there was no emotional touch and the taste of AI was still strong. They had to give up and let the colleague spend more than an hour to record it finally. The manuscript was not written by her and needed time to familiarize herself with it.

Zhang Qiao told Xinmei daybreak that, "the last colleague spent more than an hour to record it, while AI can generate it quickly, just input the text effort, and I value efficiency."

She even hoped to use AI-generated videos earlier. Currently, her team spends four hours on one two-minute video of manual cutting , without counting the collection and collation of materials beforehand. A single person can edit no more than two videos a day. Manpower is the core factor limiting the output.

Zhang Nan must understand the psychology of these creators. Before she fully devoted herself to Jianying, she spent several months doing a lot of user interviews and heard

本文链接：https://www.24zzc.com/news/171712920877108.html