OpenAI announced a new tool: it can generate videos based on the directions given by users

OpenAI has announced a new tool, named Sora, after the Japanese word for “sky”, can produce realistic images of up to a minute's duration by following a user's instructions for both subject and style, reports The guardian

OpenAI announces Sora, the new tool that can generate videos, photo Shutterstock

According to a post on the company's OpenAI blog, the model is also capable of creating a video based on a still image or augmenting existing footage with new footage.

“We teach AI to understand and simulate the physical world in motion, with the goal of training models to help humans solve problems that require interaction in the real world,” it says in the company's blog post.

One of the videos included among several of the company's initial examples is based on this prompt: “A movie trailer featuring the adventures of a 30-year-old spaceman wearing a red wool knitted motorcycle helmet, blue sky, salty desert, cinematic style, shot on 35mm film, vivid color.”

The company announced that it has currently allowed access to Sora for a few researchers and video creators. The experts will be “red team” of the product, i.e. they will test if it is able to bypass OpenAI's terms of service, which prohibit it “extreme violence, sexual content, hateful images, images of celebrities or intellectual property of others”, according to the company's blog. The company only allows limited access to researchers, visual artists and filmmakers, though CEO Sam Altman responded to user requests on Twitter after the announcement with videos he said were made by Sora. Videos carry a watermark to show they were made by AI.

OpenAI debuted the still image generator Dall-E in 2021 and the generative artificial intelligence chatbot ChatGPT in November 2022 and quickly amassed 100 million users.

For their part, Google and Meta have said they are developing video generation tools, although they have not made them public. On Wednesday, Google announced an experiment to add more memory to ChatGPT so it can retain more of its users' conversations.

OpenAI did not disclose how many images were used to train Sora, or where the training videos might have come from. He declared, however, for New York TimI'm that “thiscorpus contains videos that were both publicly available and licensed by the copyright owners'. The company has been sued several times for alleged copyright violations in training its generative artificial intelligence tools, which process huge amounts of material taken from the Internet and mimic the images or text contained in those data sets.