By reading the features of the AV1 codec and its support on sites such as YouTube and Netflix. And I came across an article on how to use it through FFmpeg, but this is not the main thing. I came across a link to a new encoder under the name "rav1e", which, by assurance, is faster than the current one.
I understand that in its current form it is not perfect. True, I would like to see in future builds.
Some time ago, I came across a Japanese project called Waifu2x. And this project is able to upscale images using a neural network from JPEG to PNG, but only through an Internet site:
Later, I found three forks using the GPU of the computer on which they are installed:
waifu2x caffe (Nvidia GPU only)
waifu2x converter cpp (for Nvidia/AMD GPUs using OpenCL instructions)
waifu2x ncnn vulkan (uses the Vulkan API driver)
But, again, a little later I came across two more forks, only now for the video:
Dandere2x (uses, optionally, three installed Waifu2x forks)
Video2x (requires the same three forks + Anime4K)
Closer to the topic. It would be nice to see this option in this plugin, as Z-Scale is not bad, but in places it loses much in performance at the processing stage.