A model which generates text in response to an input image and prompt.