Hybrid Work Had a Growth Spurt in 2023. Your Company Could Reap the Benefits

Jeff Bezos Just Sold $4 Billion Worth of Amazon Stock. But He's Still Not the World's Richest Person

Here's What Y Combinator Wants From Startups This Year

Number of Workers on Strike More Than Doubled in 2023

Department of Energy Offers $100 Million for Carbon Removal Projects

Walmart and Target Are Championing a California Ballot Measure That Could Crack Down on Retail Crime

Lyft's Typo Snafu Proves Tiny Details Can Have Big Ramifications

OpenAI has introduced Sora, its first-ever text-to-video AI model. In a series of posts on social media, OpenAI cofounder Sam Altman announced the new model and said that today, the company will begin offering access to "a limited number of creators."

According to OpenAI's website, Sora can generate videos up to a minute long in a wide variety of art styles, including hyper-realistic human faces. The model is capable of creating "complex scenes with multiple characters, specific types of motion, and accurate details of the subject and background." Sora can also take an image and generate a video from it, or take an existing video and either extend it or fill in missing frames.

Some weaknesses of the model, according to OpenAI, include that it can have trouble simulating the physics of a complex scene, not fully understanding cause and effect, and confusing spatial details in a prompt, such as mixing up left and right. So if your video features a jogger running backwards, for example, you might want to try generating it again. The model also can't create sound to accompany a video.

As for how the company plans to keep the technology from being used with malicious intent, OpenAI says they'll include C2PA metadata, which can help people to track the "origin" of the video and verify that it was created by AI, if the model is eventually deployed in an OpenAI product. If Sora is integrated into an OpenAI product, a text classifier will check prompts before they are processed to filter out any requests for "extreme violence, sexual content, hateful imagery, celebrity likeness, or the IP of others." Once the video is created, another classifier will scan it for any prohibited content.

Of the "limited number of creators" who will have first crack at Sora, they are comprised of "visual artists, designers, and filmmakers" who will give their feedback to make the model more helpful for creative professionals. AI researchers will also get access to the model for redteaming, a process in which groups find flaws and errors in the model so that it can be improved.

Altman didn't disclose when Sora would be available to the public. It has obvious potential applications for business such as creating advertisements, making content for a pitch deck, and creating social media content.

Altman solicited ideas for captions for videos from social-media users and said he would create and share them, in order to give people a sense for Sora's capabilities. His first creation took the form of "a wizard wearing a pointed hat and a blue robe with white stars casting a spell that shoots lightning from his hand and holding an old tome in his other hand."

Now accepting applications for Inc.’s Best Workplace awards. Apply by February 16 for your chance to be featured!