Sora is shutting down. Generation isn't a product
OpenAI closed the Sora app this week, seven months after launching it as a social network, and the API ends in September. What went wrong, and what it says about where generative video belongs.
The Sora app and website shut down on Sunday. OpenAI announced it in March: the consumer app closes April 26, and the Sora 2 API follows on September 24. Seven months after launch, the product that briefly topped the App Store is gone.
Reporting on why has been consistent. It was very expensive to run, since every video is a large generation, and revenue was small. Usage fell a long way from its launch peak. There were ongoing problems with copyrighted characters and deepfakes of real people. A large content deal with Disney reportedly fell apart. And OpenAI has been refocusing on coding and enterprise products.
In October I wrote a post called “Sora built a social network,” and said a feed of generations would be a great launch week and a hard long-term product, because what brings people back to video platforms is people, not production value. The shutdown doesn’t tell us that “AI video didn’t work,” though. It tells us something about this particular way of packaging it.
AI video is working. People use generation all the time: in editing apps, in ads, in short-form content, in the microdrama industry in China. What didn’t work was a destination built around generation itself.
Generation isn’t a reason to watch. The first week of Sora was people putting themselves and their friends into absurd scenes, which was fun to make and fun to share once. After that, a feed of other people’s generations had nothing to follow: no ongoing characters, no creators with a point of view you care about, no community.
Unlimited supply lowers the value of each piece. When anyone can make a polished clip in a minute, polish stops being a signal. Attention goes to the few things that have something else: a person, a story, a joke with timing, an idea.
The costs scale the wrong way. A normal video platform’s cost per view is low because content is uploaded once and watched many times. A generation-first app pays heavily for every piece of content, including all the ones nobody watches.
Rights were never solved. The most-shared content leaned heavily on characters and faces that didn’t belong to the people generating it.
What I think works instead is generation as a tool inside places people already go, attached to things they already care about. Remixing a creator’s video. Helping a creator make something they couldn’t shoot. Putting generation into editing, where it serves a person’s intent and a real audience. That’s also what I work on, so take my bias into account. The shutdown is one case; it doesn’t settle which products will work.