The publisher of the ChatGPT conversation robot and DALL-E image generator, OpenAI, has introduced a new tool called Sora that can create hyper-realistic videos that can be up to a minute long using a simple text command. Innovation in the field of artificial intelligence (AI ).
This new platform is still being tested, but that does not stop the young Californian company from demonstrating it in some videos.
Launch of the Twitter widget. Skip widget ?End of Twitter widget. Return to the top of the widget?
The program can generate videos with a maximum length of one minute while maintaining visual quality and respecting the user's wishes, OpenAI said on its website.
[Sora peut] Create complex scenes with multiple characters, specific types of movement, and precise details.
In addition to text, the tool can also create a video from a still image or even expand existing videos.
Offered to a limited number of people
Sam Altman, the head of OpenAI, announced a new tool on the social network as part of an experimental phase.
The manager also invited Internet users to make suggestions for creating videos, the most relevant of which he published on the platform a few moments later.
Launch of the Twitter widget. Skip widget ?End of Twitter widget. Return to the top of the widget?
Among these videos we can see two dogs playing in the snow in the mountains. Another video shows the flight of an imaginary animal, half duck, half dragon, against a beautiful sunset, with a hamster dressed in sportswear on its back.
[Sora sert de base à ] Programs capable of understanding and simulating the real world.
OpenAI hopes Sora will be an important step in realizing AGI, artificial general intelligence, a highly autonomous system that would outperform humans at most economically viable tasks.
Its a lot to do
OpenAI warned that the platform's current model has flaws, such as confusion between left and right or the inability to maintain visual continuity throughout the video.
For example, it could happen that a person bites into a cookie, but the cookie afterwards shows no bite marks, explains the editor.
With the presentation of this new tool, the company reiterated that the question of security represents an essential issue and that simulations would be organized with Internet users who would be asked to produce malfunctions or create inappropriate content in order to better define the limits of the platform.
“We will engage policymakers, educators and artists around the world to understand their concerns and identify positive use cases for this new technology,” said OpenAI.
Meta, Google and Runway AI, which are working on similar text-to-video applications, have also presented examples.