Neural network video generator – what Sora can do from the authors of ChatGPT – UNIAN

OpenAI has launched a neural network that generates realistic videos based on text descriptions. The development is called Sora in honor of the Japanese word for “sky”.

The main difference between Sora and similar existing models is that it generates entire videos at once, instead of combining them frame by frame. Users will be able to create videos with a resolution of up to 1080p, up to 20 seconds in length and in widescreen, vertical or square formats, writes The Verge.

OpenAI emphasized that when generating Sora, it takes into account the physical properties of objects and their interaction with each other, resulting in natural videos with natural animations. And the more specific…