I wasn't sure, as this is part of the base model and not something you edit when working on SD. He told me that it was what processed the prompt into something SD could read, and could tell you what was in an image.