OpenAI, creator of ChatGPT and the DALL-E image creator, has unveiled a new tool called ‘Sora’, which is capable of creating, by simply entering text, realistic videos that can last up to a minute – a major innovation in the field of artificial intelligence.

Based on previous research on the DALL-E and GPT programs, this new platform is still in testing, the California-based startup ally of Microsoft clarified, which did however present some videos and how they were created.

The program can create videos of a maximum length of one minute “while maintaining visual quality and respecting the user’s request,” OpenAI said on its website.

Sora can “create complex scenes with multiple faces, specific types of movements and precise details,” the start-up says on its website.

Sora also allows videos to be created from a still image, the AI ​​giant assures, or existing videos to be extended.

Sam Altman, the head of OpenAI, said on social media X that his company “will offer a limited number of creators access” to this new tool, as part of an experimental phase.

He also invited users to make suggestions for creating videos and a few minutes later uploaded the most successful ones to the platform.

Among these videos one sees two dogs playing in the snow on a mountain. Another video shows the flight of an imaginary half-duck, half-dragon animal in front of a beautiful sunset, with a hamster in an excursion outfit on its back.

Sora serves as the basis for “programs capable of understanding and simulating the real world,” explains the startup, which hopes it “will be an important milestone in the realization of AGI,” General Artificial Intelligence, a highly autonomous system that carries that it will outperform humans in most of the cost-effective jobs.

OpenAI warned that the “current model” of the platform has “flaws”, confusing left with right and showing an inability to maintain visual continuity throughout the video.

“For example, a person may bite into a cookie, but after the cookie may not have bite marks,” the company explains.

In introducing this new tool, OpenAI said that the issue of security is a significant stake, and that simulations will be organized with users being asked to produce malfunctions or create inappropriate content to better define the limits of the platform.

“We will invite policy makers, educators and artists from around the world to understand their concerns and identify positive use cases for this new technology,” OpenAI said. Meta, Google and Runway AI, which are working on similar text-to-video applications, have already shown samples.