Google has developed the Imagen Video neural network, which creates videos based on text descriptions

You've probably heard a lot about neural networks that can create images by description. The American company Google has not just decided to keep up with the trends, but has gone further.

Here's What We Know

Google has announced a neural network that can create not photos, but entire videos from text descriptions. The system of the American company is called Imagen Video. It can create video clips with a resolution of 1280 x 768 pixels. Although the duration can hardly be called long - up to 5 seconds.

Imagen Video imitates the styles of artists and can create three-dimensional figures that retain their shape during rotation. To train the neural network, developers from Google have used several tens of millions of images, videos and text descriptions.

The results are available on the Imagen Video project website. Google refuses to share the source code, arguing its decision is to prevent inappropriate content.

Source: Ars Technica