ChatGPT and Midjourney should be familiar to most people by now. But of course there are still countless artificial intelligences (and those who want to become one) that have not yet become so well known.
However, this should hardly bother the GPU developers at Nvidia. After all, they already enjoy a certain degree of notoriety. Nvidia’s Toronto AI Lab has now presented an AI projectwhich should make your pictures moveable.
Harry Potter and the Latent Diffusion Models
Latent Diffusion Models (LDM) are artificial intelligences that generate videos, without much computing power need to. According to Nvidia, work on their project is based on text-to-image generators, such as Stable diffusion. In addition, they supposedly added a “temporal dimension.”
link to imgur content
What does that mean? Put simply, this means that still images should be animated “realistically.” So a single image will supposedly become a video – more specifically, a GIF. This reminds us strongly of the moving pictures from Harry Potter. But the meme potential also seems limitless.
The project is intended to use upscaling technology to display movements that appear as real as possible in good quality. A picture should be like this 4.7 second video with a resolution of 1,280 x 2,048 pixels. With a resolution of 512 x 1,024, the videos should also be able to be longer.
link to imgur content
This means a big step in the text-to-video area and could offer various possible applications in the future, for example in the film industry.
In the current state, the quality probably leaves something to be desired, as you can still see artefacts. Also, the ever-changing environment looks very artificial at the moment, but it’s no secret that AI technologies have the property of advancing very quickly.
We remain curious to see what will happen in this area in the near future.
Are you going to use the Nvidia technologists to upgrade your meme game or do the videos still look a bit too scary for you at the moment? Some of them can get chills down your spine. Are you looking forward to further developments in the field of text-to-video? Write it to us in the comments!