OpenAI has officially unveiled its latest text-to-video and audio AI generator, Sora 2.
So far, we’re not quite convinced by the company’s marketing efforts. In an unsettling promotional clip, an algorithmically synthesized version of OpenAI’s CEO, Sam Altman, announces the company’s answer to Meta’s Vibes app — a TikTok-like experience designed to expose users to an endless deluge of AI slop.
The issue? The AI-generated Altman suffers from chronic issues plaguing AI-generated footage: it looks unconvincing and unreal.
“One year, we redefined what is possible with moving images,” the fake Altman says in a clearly robotic and AI-generated voice, referring to the company’s preceding Sora video generator, which was released in December 2024.
“Today, we’re announcing the Sora app,” he added, while a pair of piercing blue eyes stared unblinkingly into the camera.
It’s a disconcerting appearance, falling just short of clearing the uncanny valley.
The rest of the promotional video is a mind-numbing parade of AI-generated clips, from jockeys riding unicorns to a stadium-lit Altman attending what appears to be a sporting event involving humans riding oversized ducks.
Users on social media were taken aback by the slop fest — and didn’t hold back in their feedback.
“This feels uncomfortably soulless to watch,” one X-formerly-Twitter user chimed in. “It’s somehow like watching a dead person dance, very unsettling. Like the beginning of a nightmare, where you haven’t seen the monster yet, but you already know something is very wrong.”
“What the fresh f**king hell is this turd-pile?” cinematographer James Rogers added. “I hate everything about this. Unsettling, empty, soulless filth.”
“Prompting isn’t creativity,” one user tweeted. “It never will be.”
OpenAI is betting huge on its video generator, claiming that Sora 2’s “advanced world simulation capabilities” will be “critical for training AI models that deeply understand the physical world.”
In one demonstration, the company showed that Sora 2 effortlessly rendered a video of a gymnast flipping on a balance beam — a subject that led its preceding model to hilariously fail the “Turing test for AI video” by turning the athlete into a grotesque, flailing ball of far too many limbs.
While it’s certainly an improvement over last year’s AI video generator, countless users are already balking at OpenAI’s efforts to force-feed even more AI-generated slop to the masses by packaging it into a smartphone app.
“We will produce levels of slop one could only imagine before,” one X user joked.
OpenAI’s originality also leaves something to be desired. The news comes roughly a week after Meta unveiled Vibes, an extremely similar TikTok-like video feed of AI slop. Netizens weren’t impressed with that effort either, calling it “hot garbage” and an “infinite slop machine,” while questioning who could possibly want to swipe through fake videos of snowboarders jumping over rubber ducks or a cat-headed news anchor.
“Good news y’all, Mark Zuckerberg’s new idea that we’re forcing on everybody is finally here,” The Daily Show‘s Michael Kosta said in a scathing satire of Vibes.
“Our AI video tools are set to completely revolutionize how dumb f**king losers create sh**ty little videos,” Kosta added. “And we’re so stoked about it.”
In short, both OpenAI and Meta will certainly have their work cut out to convince users that an app entirely dedicated to AI slop will meaningfully justify the tens of billions of dollars they’re pouring into the tech.
“Eat your slop, piggies,” Kosta said mockingly. “Eat it from the palm of our hands.”
More on AI slop: OpenAI Launching TikTok Competitor for Short-Form AI Slop Videos