A Microsoft-backed startup is wowing social media with hyper-realistic videos created using text prompts.
ChatGPT creator OpenAI has announced a new form of artificial intelligence that creates realistic videos based on text prompts, sparking a surprising reaction online.
The text-to-video model, named Sora, has a “deep understanding of language” and can generate “compelling characters that express vivid emotions,” OpenAI said in a blog post Thursday.
“Sora can generate complex scenes with multiple characters, specific types of motion, and precise details of subjects and backgrounds,” says the Microsoft-backed startup.
“The model understands not only what the user asks for in a prompt, but also how those things exist in the physical world.”
On X, OpenAI CEO Sam Altman lets users post realistic videos of things like two golden retrievers podcasting on top of a mountain, a grandma making gnocchi, and marine animals participating in a bicycle race on the ocean. Previously, we asked users to suggest prompts about Sora. .
https://t.co/uCuhUPv51N pic.twitter.com/nej4TIwgaP
— Sam Altman (@sama) February 15, 2024
The hyper-realistic quality of the video sparked shocking reactions across social media, with users calling the results “out of this world” and a “game changer.”
“It’s been two hours and my brain still can’t process the OpenAI Sora video that was generated,” said X user Allen T.
The demonstrations fueled concerns about potential risks, especially in a year when elections are closely watched around the world, including the U.S. presidential election in November.
OpenAI said in a blog post that it will take several important safety measures before releasing Sora to the public.
“We work with red teams, experts in areas such as misinformation, hateful content, and bias, who adversarially test our models,” the company said.
“We are also building tools to help detect misleading content, including a detection classifier that lets you know when a video was generated by Sora.”
OpenAI also acknowledged that Sora has weaknesses such as continuity and difficulty distinguishing left and right.
“For example, if a person bites into a cookie, there may not be a bite mark left on the cookie afterward,” said the San Francisco-based startup.
OpenAI’s rivals Meta and Google have also demonstrated text-to-video AI techniques, but their models don’t produce as realistic results as Sora.
SORA is truly out of this world.
OpenAI’s new Text-to-Video model has been released, and it’s insane.
More examples below ⬇️ pic.twitter.com/qbMy5Rz5Mc
— Linus (●ᴗ●) (@LinusEkenstam) February 15, 2024
