ByteDance’s AI ambitions are being hampered by computing limitations and copyright concerns

move Sora 2There’s a cool new AI-powered video model in town.

In early February, ByteDance It has unveiled Seedance 2.0, a major upgrade to its flagship video model, which previously remained somewhat mysterious. Its powerful capabilities immediately shocked China’s AI ecosystem, even among audiences that were once skeptical of AI-generated videos and viewed the technology primarily as a way to produce cheesy videos.

Feng Jie, founder of Game Science, the studio that developed the globally successful Chinese video game Black Legend: WukongHe wrote online that he was “deeply shocked” by the model’s capabilities and believes Seedance 2.0 will pose major challenges to China’s current copyright regulations and content moderation systems. Pan Tianhong, who leads a Chinese professional video production studio with more than 15 million social media followers, posted a video in which he said Seedance 2.0 is far better than any video making models that came before it. “He thinks like a director,” Pan said.

However, most people cannot obtain the form at this moment because access is still somewhat restricted. As of this week, ByteDance is only allowing existing users of its consumer-facing AI apps in China — the most popular of which is Chatbot Doubao appbut the company also has a confusing array of lesser-known apps like Jimeng, Xiayunque, and Spark — for a Seedance 2.0 experience. All of these apps are intended for the Chinese domestic market only, preventing people outside the country from testing the model for themselves. (The restrictions have prompted some smart people in China to resell their ByteDance accounts to eager AI early adopters abroad.)

But there are signs that this model may soon become more available. This week ByteDance updated its API platform and revealed the proposed price for Seedance 2.0: a 15-second video, the longest video that can be produced at the moment, will cost just over $2, according to the Chinese publication. Estimated IT major. ByteDance still hasn’t opened up access to its API to third-party developers, but that should be on the horizon.

Aphra Wang, Substack Newsletter Author synchronous Seedance 2.0 is another interesting example of how the two countries are taking diverging paths, one close observer of the US-China AI landscape tells me. Even before Seedance 2.0, some of the world’s most established AI tools for video making, such as Kling AI, It is developed by Chinese companies. “China has not produced any suitable tool for encrypting AI, which is why Chinese people all rely on Cloud Code or Codex,” says Wang. “But when it comes to video AI, China is miles ahead of the United States.”

But all the hype aside, the Sedan has two serious problems. Weeks after its launch, ByteDance is experiencing a computational bottleneck that causes the model to take hours to generate a single video. Meanwhile, major film studios, including Disney, Netflix and Paramount, have sent cease-and-desist letters to ByteDance alleging that Seedance 2.0 output infringes their copyrighted works. ByteDance did not immediately respond to a request for comment.

Bandwidth problem

Even if you have access to the ByteDance AI app, it’s still difficult to make a video with Seedance 2.0, because a lot of people are trying to do the same thing, and ByteDance hasn’t yet made enough computing resources available for everyone.

When I tried to create a clip using one of the ByteDance apps this week, it told me I was number 90,985 on the queue, and it would take about four hours to create a five-second video. After waiting for two hours, the app told me I now had six more hours left. At that point, I decided to just go to sleep.