Google presented on Thursday (25) Lumiere, a artificial intelligence (AI) model capable of generating realistic videos from textual descriptions. Developed in partnership with Tel Aviv University and the Weizmann Institute of Science, both in Israel, the technology can also transform static images into animated videos.
Described by the creators as a “text-to-video diffusion model designed to synthesize videos that depict realistic, diverse, and coherent motion,” the new AI is based on the Space-Time U-Net architecture. The structure generates all the frames that make up the video at onceunlike similar mechanisms.
According to the company, this leads to greater cohesion and precision in movements, providing more realistic videos, even at a lower resolution. In general, just type the text commands or insert the desired photos and wait for the tool to do its work.
Another difference with Google Lumiere is the possibility of generate videos in specific styles. The user can choose to animate just one part of the image, for example, while the rest remain static, generate new content while maintaining the same visual pattern as the original photo, modify object characteristics and much more.
Risks of misuse
In the article introducing Google's new AI that generates realistic videos based on text and photos, the developers state that the objective is offer novice users the chance to create visual content in a creative and flexible way. However, they mentioned the risk of misuse of the technology.
Regarding the possibility of taking advantage of the tool to create fake videos, the team said it was It is necessary to develop ways to detect “cases of malicious use” of the resource to ensure everyone’s safety. Concern about the spread of fake news becomes even greater in an election year.
Google just made an incredible AI video breakthrough with its latest diffusion model, Lumiere.
2024 is going to be a massive year for AI video, mark my words.
Here's what separates Lumiere from other AI video models: pic.twitter.com/PulSjVZaCp
— Rowan Cheung (@rowancheung) January 25, 2024
For now, Google Lumiere is not available to the general public and there is no forecast for when it will be offered. Rumors suggest that the new AI-powered video generator could be integrated with Google Bard, although big tech has made no mention of this.