Google's new AI, Lumiere, creates short videos based on text or even an image provided to it as an example.

Google has developed a new artificial intelligence. It is no coincidence that the device called Lumière bears the name of the Lumière brothers, who invented the technology of recording and displaying moving images, as the algorithm can be used to create video clips, according to what was reported. Ars Technica.

Based on the video posted by the company, the system creates realistic recordings lasting a few seconds using text commands. He can do all this even if we show him a perfect picture, and he creates the recording in his style.

See also  The Japanese lunar rover's mission may be over forever, but it has sent one last image

But according to Google, it can also be used for video editing, for example, to replace the clothes worn by one character with another, or to turn a specific part of a still image into a video. For the latter, think, for example, of a photographed campfire: with a few clicks, you can select the desired area, which is then animated by Lumiere.





The truly groundbreaking solution is for the algorithm to deal with space and time at the same time. In the former, the algorithm monitors where certain objects are located in the video, and in the latter, how the subject moves and changes over time.

This is also why the result is so realistic.

Google hasn't said where the amount of data needed to train the model came from, but it claims to have used 30 million videos and the accompanying text prompts it to train Lumiere.

Videos are 80 frames long and play at 16 fps (frames per second). In other words, the result is 5 seconds of content.

It's not yet known when field users will be able to use the tool, but we wouldn't be surprised if the company is the first to integrate Lumiere into YouTube.

If you want to know about similar things at other times, like it HVG Tech Department Facebook page.