Does vision solve it? How does one ask a question through images alone?