Microsoft-backed startup stuns social media with hyper-realistic videos created using text prompts.

OpenAI, the creator of ChatGPT, has unveiled a new form of artificial intelligence that creates realistic video based on text prompts, prompting stunned reactions online.

The text-to-video model, named Sora, has “a deep understanding of language” and can generate “compelling characters that express vibrant emotions,” OpenAI said in a blog post on Thursday.

“Sora is able to generate complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background,” the Microsoft-backed startup said.

“The model understands not only what the user has asked for in the prompt, but also how those things exist in the physical world.”

OpenAI CEO Sam Altman on X invited users to suggest prompts for Sora before posting results that included realistic videos of two golden retrievers podcasting on top of a mountain, a grandmother making gnocchi, and marine animals taking part in a bicycle race on top of the ocean.

The hyper-realistic quality of videos prompted stunned 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 these generated OpenAI Sora videos,” X user Allen T said.

The demonstration also promoted concerns about potential risks, especially in a year of closely watched elections around the world, including the US presidential election in November.

OpenAI said in its blog post that it would be taking several important safety steps before releasing Sora to the general public.

“We are working with red teamers  –  domain experts in areas like misinformation, hateful content, and bias  - who will be adversarially testing the model,” the company said.

“We’re also building tools to help detect misleading content such as a detection classifier that can tell when a video was generated by Sora.”

OpenAI also acknowledged that Sora has weaknesses, including difficulty with continuity and distinguishing left from right.

“For example, a person might take a bite out of a cookie, but afterward, the cookie may not have a bite mark,” the San Francisco-based startup said.

OpenAI rivals Meta and Google have also demonstrated text-to-video AI technology, but their models have not produced results as realistic as Sora’s.

 

Comments are closed.