This seems like it would really benefit from a model trained specifically to upscale video. There's a lot of information you can get from the structure between

This seems like it would really benefit from a model trained specifically to upscale video. There's a lot of information you can get from the structure between frames that you can't exploit with a still image.