The race for innovation in the field of artificial intelligence continues at a breakneck pace. Google presents Lumiere, an AI capable of generating videos from an illustration provided by the user or generated on the spot by the same artificial intelligence. If AI of this type already exists, Google here promises great realism in movements thanks to a new mechanism for processing spatio-temporal scales.
Generate videos using artificial intelligence
Barely announced, the technology is already making a big splash in the tech sphere. Google’s AI is in fact capable of generating a video from a still image and even a prompt entered in natural language. A second feature currently in development aims to create, based on the style of an image, several videos with different thematic registers.
Google reveals the capabilities of its Lumiere AI in this video published on YouTube. We see animations created from scratch, as well as an illustration of the concept of generating multiple videos based on the style of a reference image.
The Mountain View firm explains that this new artificial intelligence is based on a new mechanism responsible for processing spatio-temporal scales, offering renderings with realistic movements for the human eye.
Note that Google is not the first to launch artificial intelligence specific to video generation. An online tool already available does the same, but in a visibly less complete and less realistic way in terms of movements. This is Leonardo AI, which we already mentioned in our columns a few weeks ago.