Well, I will start by clarifying that the objective of the NVIDIA researchers who have participated in this project has not been specifically stated in this way, that is, no one came up with the idea of creating a generative artificial intelligence model specifically designed to create animated GIFs, that aim to become memes on social networks. However, and as you will see in this news item, the truth is that this is exactly what has come out of his artificial intelligence laboratory in Toronto, Canada, to the special happiness of all those who are already considering the thousand and one a memes they want to create.
It is well known that NVIDIA already put the focus, a few years ago, on artificial intelligenceand that they have done so both from the hardware perspective, designing and producing chips specially equipped for the requirements of this discipline, and from the software perspective, a field in which they have managed to become a benchmark with the most diverse developments, ranging from its DLSS technology to applications like GauGAN2, to many other successful experiments and proofs of concept that extend the reach of artificial intelligence.
If you are already familiar with GauGAN2, the technology we have talked about before, you will better understand how High-Resolution Video Synthesis with Latent Diffusion Models works, which translated becomes High-Resolution Video Synthesis with Latent Diffusion Models. It is a model that, like GauGAN2, works with text inputs, but that uses the technological base of Stable Diffusion to generate short, high-resolution videos. More specifically, this model can generate videos of up to 4.7 seconds with a resolution that can go up to 1,280 x 2,048 points. In case of reducing the resolution, it is possible to obtain longer videos.
To achieve this, this model combines the capacity of Stable Diffusion with the nature of latent diffusion models, which, as you can see in this case, allow a specific image to be given temporal dimensionality, mode in which an animation is obtained with a prompt similar to the one we would use in an image generation AI, but in which we will also have to include the details related to the development of its action.
You can check the result in the many examples published by NVIDIA on the model information web page, where you can also find a lot of technical information about its operation. In these examples you will see that we are not talking, at the moment, about videos that could be considered totally real (although the driving simulations are amazing), but we must bear in mind that we are talking about an incipient development, so it isYour capacity for evolution is very high. And, in the meantime, it is worth starting to think about precisely what I was saying at the beginning, that is, about the immense number of possibilities that a model like this offers when creating animated GIFS that can, already as memes, end up turning around to the world.
