Google has unveiled a promising new artificial intelligence tool. It is called Lumiere and is defined as “a spatiotemporal diffusion model for video generation.”
Technological advances continue to revolve around artificial intelligence and video generation. Following the recent launch of the GPT Store's " Videomaker " tool, which allows users to create videos from detailed text prompts using AI, Google is now introducing what promises to be the most advanced AI video generator to date .
If you cannot view the video correctly, click here .
How does this new tool work?
From adding glasses to an owl or putting boots on a moving duck to creating a video of the Mona Lisa laughing out loud , Lumiere can generate videos from a still image or text, but it also offers other possibilities . For example, animate only part of a user-supplied image, such as making a butterfly's wings move while keeping the rest of the landscape static; or use an image alone to indicate the design style or colour you want in your video.
Lumiere
To implement this tool, they lebanon number screening have used an architecture called Space-Time U-Net that generates the entire duration of the video in a single process. “Our model learns to directly generate a low-resolution, full-frame video by processing it at multiple space-time scales, using spatial and temporal down-sampling and up-sampling, and taking advantage of a previously trained text-to-image diffusion model,” explains Google .
Although promising, Lumiere still has some limitations: its videos are no longer than 5 seconds and do not reach optimal quality. However, considering the giant steps that artificial intelligence has made in terms of video generation in just one year, everything points to it being only a matter of time before this tool improves.
- Board index
- All times are UTC
- Delete cookies
- Contact us