Just over half a month in, OpenAI must honestly face an awkward reality: Sora2, the video generation platform they once had high hopes for, is slowly turning into a flashy junk-video generator.
When Sam Altman promised the world a revolution in the video field, he probably didn't expect that what would ultimately be delivered was just a high-end toy that users would get tired of after five minutes of novelty.
When Sora2 first appeared, it amazed everyone. With just a piece of text or an image, it could generate surreal short films comparable to actual footage, igniting the dream in countless people of becoming a great director. The media frenzy and the tech circle's hype further pushed it onto a pedestal.
However, as excited users flocked to the platform, the beautiful dream quickly collided with cold, hard reality.
In fact, the so-called intelligent creation is just a frustrating gacha game. Basketballs hang weirdly in the air, characters' clothes change color seamlessly between shots, and text renders into a pile of gibberish, not to mention the stiff facial expressions and awkward speaking voices.
This is not filmmaking at all; this is a long war of wits against the randomness and bugs of AI.
The professional film production team Shy Kids revealed that they need 300 generations to get a few usable clips, a number that means its creative efficiency is hundreds of times lower than the most traditional manual shooting.
Other users also quickly discovered that Sora2 only gave them a random button full of bugs. It cannot stably realize your creativity; it can only randomly reward you with some bizarre and motley clips.
I. The Dream of an AI-version of TikTok
If the birth of Sora2 was a splendid firework, then its rapid silence stems from three unforgivable sins in its product design. When we compare it horizontally with the true short video hegemons like TikTok and YouTube Shorts, the shattering of this dream becomes incredibly clear.
1. How to Retain Users with Ephemeral Content?
The content of mainstream short video platforms is rooted in people, a kaleidoscope of real life. Whether it's street interviews on Douyin, rural life on Kuaishou, or talent challenges on Reels, the core is always lively people and things that can resonate with others.
Sora2's content, on the other hand, is 100% synthesized by AI, essentially a visual spectacle show. Users might be amazed at first by a cyberpunk-style "Along the River During the Qingming Festival," but this amazement is short-lived. When the threshold for visual stimulation is constantly raised, what remains is emptiness and boredom. Because it lacks the core anchor of "people," the content loses its ability to continuously generate emotional connections.
As one netizen said: "On Douyin, I can see a hundred kinds of real life; on Sora2, I can only see one kind of fake excitement." In the end, the space for Sora2 users to be creative is severely compressed, making it naturally difficult to produce diverse and viral content.
2. A Community Without Social Connections is Just a Video Website
Social is the underlying operating system of all content platforms. TikTok not only has algorithmic recommendations but also a complex social network based on follows, friends, comments, and challenges, which constitutes strong user stickiness.
Sora2 is almost a desert in this regard. Since the content is detached from real-life personalities, the connections between users are very fragile. You might like a video, but it's hard to follow a person behind a virtual work. The so-called follow system is also a mere formality due to the high homogeneity of the content.
A community without strong social connections is essentially just a read-only video website, not an interactive social platform. Users on Sora2 are lonely; they are just passive consumers of information, not co-builders of the community.
As the novelty of the new AI tool fades, this lonely carnival naturally comes to an end.
3. No Bread, Who Will Generate Power for You?
This is perhaps the heaviest straw that broke Sora2's back.
Platforms like YouTube, TikTok, and Instagram have long established mature creator economy ecosystems: ad revenue sharing, creator funds, live streaming rewards, e-commerce... Creators' talents can be directly exchanged for tangible income, which drives a positive feedback loop.
In contrast, Sora2 is more like a "power-for-love" club. Creators here have no ad revenue sharing, no tipping tools, let alone commercial monetization. OpenAI's business model is still API fees for B2B and subscription fees for B2C. The free labor of creators merely provides nutrients for the model's evolution and earns eyeballs for OpenAI.
A platform that cannot let creators make money is like an engine without fuel. No matter how exquisitely designed, it will eventually become a pile of cold, useless iron.
II. Why Are AI Giants Obsessed with the Pipe Dream of a Content Ecosystem?
Sora2's defeat is not an isolated case. From text to video, we have seen too many AI companies trying to build a closed content platform driven purely by AI. However, whether it's Pika, Kling, Dreamina, or countless other fleeting applications, none have truly succeeded on this path.
Behind this is the original sin of AI content that cannot be fixed by technological iteration: it lacks the soul of human creation—real emotions, unique personal experiences, and groundbreaking originality. After a brief period of curiosity, users are faced with a cold, repetitive, and unrelatable algorithmic spectacle.
Trying to build a social platform with pure AI content is like trying to build a city in the desert without a water source—it is doomed to fail.
The lesson of Sora2 is also a question for the strategic direction of the entire industry. For current AI companies, instead of spending huge sums of money chasing a pipe dream of platform traffic, it is better to be a super AI tool in a down-to-earth manner.
On this point, Google's strategy appears more sober and pragmatic. It did not try to use Gemini to create an AI version of Quora or an AI version of YouTube out of thin air. Instead, it integrated its powerful AI capabilities as an underlying infrastructure into core products such as search and workspace suites, empowering billions of users worldwide. This is the duty of a tool, and it is also the role that AI should play at this stage.
Before envisioning the endgame of artificial general intelligence, it is better to be down-to-earth, return to the essence of a tool, and think clearly about how to solve specific problems for each specific person.
Top comments (0)