OpenAI introduced a new neural network called Sora that can generate entire videos based on a text description. And it looks like another revolution in the development of artificial intelligence.

Sora is a diffusion model capable of “rendering complex scenes with multiple characters, specific movement types, and accurate object and background details.” The project website says the model can understand cues and know how different objects in the physical world behave.

The important thing is that Sora renders the entire video in one go, rather than frame by frame. This helps avoid a problem seen with other approaches (changing the subject while moving or moving the camera).

Sora is still in development, and OpenAI has no plans to make it widely available yet. Especially due to security concerns. The company is currently working on ideas to reduce hate content and clearly label AI-generated videos, among other things.

But OpenAI is already giving Sora access to a number of artists, designers, and filmmakers to get feedback on how the model can be improved to make it most creatively useful.

Separately, it is worth saying that the news was published on the OpenAI website literally half an hour ago, many videos are not yet available on other platforms. Therefore, it is worth going to the site and watching other videos created by Sora – there are a lot of them.