Image ingestion

If i upload an image, does Marqo generate the metadata needed for the vector? if so can U select which model to use, such as GPT 4.x? Or do I need to supply the metadata?