Generative AI continues to evolve. OpenAI has just launched Sora, a text-to-video model that converts sentences into 60-second videos, with a result that is both stunning and disturbing.
How far will generative AI go? A year ago, many made fun of this AI which tried to reproduce the actor Will Smith eating pasta. Today, the time is no longer for mockery. On February 15, 2024, OpenAI launched Sora, a text-to-video model whose first results are already impressive.
As always, the principle is quite simple: you give instructions to Sora and the AI takes care of converting them into a clip, the duration of which can be up to 60 seconds. OpenAI, which has already revolutionized the generative AI sector with ChatGPT, promises scenes “ imaginative and realistic “. Something to worry about artists who normally take several hours to produce this kind of content.
Sora is OpenAI’s powerful new tool
“ Sora can create videos up to 60 seconds long with detailed scenes, complex camera movements, and multiple characters with vibrant emotions ”, says OpenAI.
In this tweet published on February 15, the company gives the following prompt: “ The beautiful snowy city of Tokyo is restless. The camera moves through the busy streets, following several characters enjoying the winter weather and going shopping. Sakura petals fly in the air, accompanied by snowflakes. ” Here’s the result :
It really feels like it when you look at this sequence, with realistic animations, total respect for the scenario (it lacks petals and flakes, if you want to quibble) and a multitude of details. OpenAI provides other equally stunning examples, with certain clips that are reminiscent of a real animated film worthy of Pixar (this cute little monster).
There are still some glitches, like legs swapping in a video showing a well-dressed woman walking the streets of Tokyo in the middle of the night (the problem is visible at the 15th second). They are assumed by OpenAI: “ The current model has weaknesses. It may encounter difficulties in trying to accurately simulate the physics of a complex scene, and may not understand specific cases of causality. For example, a character can eat a piece of cookie and the next moment the cookie remains intact. »
At the moment, Sora is only available to Red members, experts responsible for assessing risks. “ We will also ensure access to a few artists, designers and filmmakers to get feedback on how to move this model in the best direction for creative professionals. »
Subscribe to Numerama on Google News so you don’t miss any news!