On the Image to Text action, there is a field to enter instructions. That is very useful:
But when it comes to a document, this field is not present, leading to a lack of precision when extracting content from a PDF:
