Of course, it was only a matter of time before the Stable Diffusion team launched its own AI model for moving images and that day has now arrived.
Stable Video Diffusion (SVD) Image-to-Video is a Latent Video Diffusion Model (Video-LDM) that can generate a short video clip from a still image as input.
Two SVD models, which are primarily intended for research purposes, were presented at the launch of Hugging Face. They have not yet been trimmed for photorealism. In contrast to closed, commercial models such as RunwayML, the Stability AI models can be freely viewed and used by anyone.
Stable Video Diffusion - Free AI diffusion model for video generation
If you would like to find out more about the details of the model, you can find a 
corresponding PDF and direct access to the code on Github here.
The two models themselves are on Huggin Face here and here
The main technical difference between the two models is the number of FPS generated (14 and 25 respectively) at a resolution of 576 x 1024 pixels.
Stable Video Diffusion - Free AI diffusion model for video generation
According to initial information, a GPU with at least 40 GB of VRAM is required in order to install and use the new models, but there are already voices on the net claiming that results can also be achieved with 24 GB. However, it is not yet clear to what extent the resolution and frame rates must/can be restricted for this. The weights themselves "only" have a size of approx. 10 GB. In this respect, there is still hope that it will be possible to experiment with consumer GPUs on your own PC in the near future with appropriate waiting times.
;)
;)