I would try this but the downside is GPT-4 vision currently doesn’t like to extract large text blocks. You could try extracting line bounding boxes with PyMupdf

I would try this but the downside is GPT-4 vision currently doesn’t like to extract large text blocks. You could try extracting line bounding boxes with PyMupdf and feeding it individual lines.