Building on the success of its image- and text-generating software, OpenAI has unveiled a new product that can generate short videos from simple text prompts and still images.
The new tool, Sora, can create high-quality videos up to a minute long, OpenAI wrote in a blog post announcing the new product on Thursday.
"We’re teaching AI to understand and simulate the physical world in motion, with the goal of training models that help people solve problems that require real-world interaction," OpenAI wrote.
Sora is far from the first generative AI video-creation tool, but OpenAI is considered one of the most advanced developers of the large language models required for such a product.
It's also the first new product OpenAI has released in more than a year, and only a few months since the former board's failed attempt to oust CEO Sam Altman.
The tools is not broadly available to the general public — yet.
"Today, Sora is becoming available to red teamers to assess critical areas for harms or risks. We are also granting access to a number of visual artists, designers, and filmmakers to gain feedback on how to advance the model to be most helpful for creative professionals," OpenAI wrote.
Red teamers are trusted actors who attack a product or service with an organization's permission for the purposes of finding cracks in the system.
In OpenAI's case, they will be "domain experts in areas like misinformation, hateful content, and bias — who will be adversarially testing the model," the company wrote.
Already, deepfakes, or videos modified or generated to deceptively portray a real person, are a concern in politics and culture, and AI technology promises to speed their creation and quality. Congress is considering what to do about the problem — including examining the body's own use of AI. And the use of AI in video creation was a major issue in last year's Hollywood strikes.
OpenAI presented more innocuous examples of Sora at work, sharing dozens of examples of videos that it says were all "generated directly by Sora without modification" through a process called a diffusion model.
A video created through that process initially appears as "static noise." That noise is gradually removed as it transforms into a finished video.
The technology can also be used to animate still images, in addition to extending or filling in existing video content, OpenAI says.
One of the many videos shared by OpenAI features "a stylish woman" walking "confidently and casually" in Tokyo surrounded by "warm glowing neon and animated city signage," as the 64-word prompt describes.
A different 15-second-long video shows a "photorealistic closeup video of two pirate ships battling each other as they sail inside a cup of coffee."
As part of its stated safety measures, OpenAI wrote that it's "also building tools to help detect misleading content such as a detection classifier that can tell when a video was generated by Sora."
Altman shared the news on X/Twitter, prompting reactions from others.
"Sam plz don’t make me homeless," well-known YouTuber MrBeast wrote in a reply to Altman.
MrBeast, who is also known as Jimmy Donaldson, has built a massive online following and fortune by creating a content and business empire. The 25-year-old video creator has 239 million subscribers on YouTube.
"Will generate you a video, what would you like?" Altman posted in response to Donaldson's joking plea.
"Hmmm, a monkey playing chess in a park," Donaldson replied.
A little over two hours later, Altman delivered a 10-second video of a monkey wearing a plain red hat sitting on what appears to be a long wooden table in a park. In front of the monkey is a chess board with three black and four white chess pieces.
