With Sora, the generation of videos through artificial intelligence can reach a milestone.

A woman walking the streets of Tokyo under the glow of neon lights, a bird’s-eye view of a lighthouse in the middle of a cliff, a crab fighting an octopus: these videos are so believable that we can take them for real. However, by typing simple text instructions into an artificial intelligence (AI) called Sora, they are generated without a camera, 15 February

By OpenAI, publisher of ChatGPT.

As fascinating as it is, these images should be viewed with caution: the media and the general public have not yet been able to test Sora. Little is known about the conditions under which these example videos were made. For example, we know nothing about the proportion of failed sequences or the number of trials with common visual oddities with this type of apparatus, such as a two-legged horse or a three-legged man.

Also, Sora is not the first video generation tool of its kind. Meta unveiled Make-a-Video in 2022, the same year as Google’s Imagen, which now has a successor called Lumiere. Since 2023, commercial instruments have appeared such as Runway ML, Pika or Kaiber. Many AIs with disappointing results according to an overview conducted by world In June 2023.

less stable

Sora’s videos look better in more ways than one: more detailed, more realistic, longer. According to OpenAI, they can last up to a minute while many competitors limit their sequences to a few seconds. Above all, Sora’s videos are far less static than the AI ​​we’re used to, often limited to more or less animated 2D scenes.

In comparison, Sora’s demonstration sequences are populated by elegant camera movements, sometimes carried on a shoulder, sometimes mounted on a drone, or even moved on rails with the geometric inertia of a video game. These beautiful cinematic qualities will not fail to make video professionals question: From what image base did OpenAI draw these videos to train Sora’s artificial neurons? In its research report, the American company continues to ignore On this topic. she declares New York Times Use royalty-free videos as well as copyrighted images.

Also Read | Articles are reserved for our subscribers On Instagram, artists using AI search for meaning: “These images are not entirely mine”

Add to your preferences

It is not clear when Sora will be open to the general public. Access to the new OpenAI tool is currently reserved for hand-picked image professionals as well as researchers who want to help lock AI down against malicious uses – specifically against content generation. Violent, obscene images or misinformation campaigns.

You have 40.03% of this article left to read. The rest is reserved for subscribers.