An anonymous reader cites a report from Wired. We already know that OpenAI chatbots can help you pass the bar exam without going to law school. Now, just in time for the Oscars, his new OpenAI app called Sora wants you to master film without going to film school. Currently a research product, Sora will work with a few select authors and a number of security experts who will red team on safety vulnerabilities. OpenAI plans to make this available to all aspiring writers on an unspecified date, but we decided to preview it in advance. Other companies, from giants like Google to startups like Runway, have already unveiled text-to-video AI projects. But OpenAI says Sora is known for its impressive photorealism (something we haven't seen from competitors) and the ability to create longer clips (up to a minute) than the short snippets other models typically do. It is said that it is a characteristic. The researchers I spoke to wouldn't say how long it takes to render all that video, but when asked, they said they'd rather eat a burrito than take a few days off. He explained that it was something like “I'm going to go to.” If the examples I've selected are to be believed, it's worth the effort.
At OpenAI, we shared four instances of Sora's power, even though we couldn't type in the prompts ourselves. (Nothing came close to his supposed one-minute time limit; the longest was his 17 seconds.) The first message was from a detailed prompt that sounded like the setup of an obsessive screenwriter. It was something. Several people were seen enjoying the beautiful snowy weather and shopping at nearby food stalls. Gorgeous cherry blossom petals are flying in the wind along with snowflakes. ” The result is a magical moment where snowflakes and cherry blossoms coexist, an unmistakable and convincing view of Tokyo. A virtual camera that looks like it's attached to a drone follows a couple as they take a leisurely stroll through the cityscape. One passerby is wearing a mask. Cars roar past on the riverside road to the left, and shoppers weave in and out of small shops on the right.
It's not perfect. It's only after watching this clip several times that I realize that if the virtual camera had continued working, the main couple strolling down the snow-covered sidewalk would have faced a dilemma. The sidewalk they occupy appears to be a dead end. They would have had to climb over a small guardrail and emerge onto a strange parallel walkway on the right. Despite this minor glitch, Tokyo Example is an amazing exercise in world-building. In the future, production designers will be debating whether it's a powerful collaborator or a job killer. Additionally, the people in this video are generated entirely by digital neural networks, are not shown in close-up, and do not express any emotion. However, the Sora team says there have been other instances where fake actors have expressed real emotions. “It will be a very long time before text-to-video conversion threatens actual film production,” he concludes for Wired. “No, you can't stitch together 120 one-minute Sora clips to create a consistent movie. Continuity is impossible because the models don't respond to prompts in exactly the same way. But time There are no limits.'' This is a barrier for Sora and programs like it to transform TikTok, Reels, and other social platforms. ”
“To make a professional film, you need very expensive equipment,” says Bill Peebles, another researcher on the project. “This model allows the average person who creates videos on social media to create very high-quality content.”
References: OpenAI develops web search product to compete with Google
