Artificial intelligence company OpenAI on Monday released its video generation program Sora for customers to use.
This program takes written prompts and creates digital videos of up to 20 seconds.
ChatGPT’s creators launched a beta version of the program in February and released a general version of Sora as a standalone product.
“We don’t want the world to be all about text,” OpenAI CEO Sam Altman said in a livestreamed announcement Monday. , I think we’re missing something important.”
In a blog post announcing the general release, the company said it wants to be at the forefront of creating culture and rules around the use of AI-generated video.
Holiday Sales: Shop this season’s hottest products and sales, handpicked by our editors.
“We are now introducing video generation technology, which allows society to explore its potential and develop norms and safeguards to ensure that it is used responsibly as the field advances.” “This is to give us time for joint development,” the company said.
What can Sora do?
The program uses a “deep understanding of language” to interpret prompts and create videos with “complex scenes” of up to one minute, including multiple characters and camera shots, specific types of movement, and precise details. Create.
Examples OpenAI gave during its beta launch ranged from animated monsters and kangaroos to videos of real people, such as a woman walking the streets of Tokyo and a movie trailer showing an astronaut in a salt desert.
The company said in a blog post that the program still has limitations.
“It often generates unrealistic physics and struggles with complex movements over long periods of time,” the company said.
OpenAI announces protection from abuse
Critics of artificial intelligence point to the potential for the technology to be misused, with real-world examples including a deepfake of President Joe Biden telling voters not to vote and a blatant AI-generated Taylor・It cites incidents such as the deepfake photo of Swift.
OpenAI said in a blog post that it will restrict uploads of people, but will loosen the restrictions as the company improves its deepfake mitigations.
“Our top priority is to protect particularly harmful forms of abuse, such as child sexual abuse material (CSAM) and sexual deepfakes, by blocking their creation, filtering and monitoring uploads, and using advanced detection tools. “If you are identified as a child at risk, you can prevent this by filing a report with the National Center for Missing and Exploited Persons (CSAM) or Children at Risk (NCMEC),” the company said.
OpenAI said all videos created by Sora will include C2PA metadata and watermarks as default settings to allow users to identify videos created by the program.
