Original video in Chinese.
Key Takwaway
- Sora 2 released: OpenAI launched the Sora 2 model and app during the holidays, with realism and physical accuracy reaching a new high; I tested it by generating a video that got strong views, and viewers couldn’t tell it was AI.
- Appcreative format: Users upload real facial video material, set permissions, and let AI generate performances in virtual scenes; mixing reality and imagination, driven by humor, it represents the unique fun of AI dismantling reality.
- Generational impact: AI Douyin is more wild and fresher, and everyone can create; older people are confused and left behind. The app is only available in the US and Canada and requires an invite code (cheap on Xianyu), and I recommend jumping in quickly, like the ChatGPT moment for AI video.
The first to make an AI version of Douyin was not ByteDance, but OpenAI.
That scheming Sam Altman caused trouble again during the holidays. The night before yesterday, Sora 2 was released, along with a dedicated app.
To test just how powerful Sora 2 really is, I used it to generate a video. The result got 100,000 views on Video Account and 270,000 views on Douyin.
One decisive factor behind this video’s traffic was that it was realistic enough. If you look at the comments, a lot of viewers simply couldn’t tell it was AI-generated.
That realism comes from the model. Sora 2 has raised the bar to the current highest level in realism. The characters’ expressions, lip movements, voices, and actions all work together extremely well.
What’s even scarier is that Sora 2 has also made huge improvements in physical accuracy. Those complex physical interactions were almost unsolvable for previous models, but Sora 2 can simulate them perfectly now.
That means AI’s understanding of the real world has taken another step forward, and we’re one step closer to a world model.
Beyond the model layer, OpenAI has also put a lot of effort into the application layer. Don’t be fooled by how rough the app looks right now; the corecreative format is already very clear:
Anyone can use the front-facing camera to record a short video, read out three numbers, and turn their head a few times, then upload it to AI as source material. After that, you set permissions for who can use your material to generate content. For example, Sam Altman opened the permissions to everyone. So over the past couple of days, my timeline has been nothing but him.
This is the uniquecreative format of AI Douyin: a large number of real people and virtual scenes mixed together, performing under the drive of humor and imagination.
This is the unique pleasure of AI’s generation: completely deconstructing reality and totally messing it up.
These new things represent their own taste and a distinct identity. Other people just don’t get it, and can’t participate.
Especially the older crowd—they’re arrogant and totally confused. I laughed when I saw the investors’ comments. They said that they can’t even keep up with the real Douyin, so who would watch AI Douyin?
The problem is, AI Douyin content is wilder, fresher, and more fun. And as long as I want to, as long as I have taste, I can generate it too.
In the past, 1% of people created content for 99% of people to watch. Now anyone can be both a consumer and a producer. I’d even say that in AI Douyin, everyone is a producer and a creator.
When generational change happens, it never transitions gently. It abandons the old brutally and completely. That is what all the big shots fear most: being eliminated without even understanding what’s happening. That is the most painful, and also the most humiliating, thing.
Right now, the Sora app is only open to the US and Canada, and it requires an invite code. But that’s fine—there are already tons of them for sale on Xianyu. The price has dropped from the original 500 to just a few yuan. I bought mine for 9.9 yuan. I recommend everyone go get one quickly too. Because this moment is the ChatGPT moment for AI video.
OK, that’s all for this episode. If you want to understand AI, want to become a super individual, and want to find like-minded people, come join our NewType community. See you next time!