Currently, Kamifusen only handles images, creating variants. We should be able to handle all representable blobs (including PDFs and videos). Using `.representation(options)` is more open than a `.variant(options)` or a `.preview(options)`