Google wants to take the lead in the field of artificial intelligence. The company’s researchers have unveiled their plans for a new spatio-temporal diffusion model called Lumiere. What is he doing?

This AI model It has the ability to bring words and static images to life and convert them into video sequences. Lumiere is designed to represent “realistic, diverse and coherent movements” through what it calls the “U-Net Space-Time Architecture.”

Google enables electronic signatures in Drive and Docs in Workspace

While it is not the first AI model of its kind, it is the first to produce realistic results with smooth movements. Well, until now, videos made with AI have left a lot to be desired and are easily recognizable.

The system generates “the entire temporal duration of the video at once, through a single pass in the model.” This means that Lumiere does not synthesize the different frames separately and then project them consecutively, but processes the sequence as a whole, resulting in a video with smooth and realistic movements.

This was the most searched for movie on Google in 2023 and was also a success at the box office

Although Lumiere is not yet available to the general public, expectations about its capabilities are high. It is also expected to bring static photos to life or can complete videos where parts have been deleted or cut prematurely.