Some visual frame-by-frame input experiments with VQGAN in 2020-2021