Not even filters; every text2image model ever created thusfar, can be very easily nudged with a few keywords into generating outputs in a specific visual style

Not even filters; every text2image model ever created thusfar, can be very easily nudged with a few keywords into generating outputs in a specific visual style (e.g. artwork matching the signature style of any artist it has seen the some works from.) This isn't an intentional "feature" of these models; rather, it's kind of an inherent part of how such models work — they learn associations between tokens and structural details of images. Artists' names are tokens like any other, and artists' styles are structural details like any other. So, unless the architecture and training of this model are very unusual, it's gonna at least be able to give you something that looks like e.g. a "pencil illustration."