The quotas for summarising text and parsing images and then summarising text aren't the same. As you surely know.
Who’s paying for that? Certainly not the users (yet).