What is pdf2md?

PDF2MD is a Python tool that batch converts PDF technical documents, such as chip datasheets, reference manuals, and spec sheets, into clean Markdown files. It is built for engineers who need to search, compare, or feed these documents into an AI model, a task that plain text extraction handles poorly. The README explains that an existing tool, Microsoft's MarkItDown, does a reasonable job with basic text but falls short in three specific ways: it extracts almost no embedded images, it produces garbled output when a confidentiality watermark overlaps with the real text, and it misses borderless register tables that are common in chip documentation from ARM and Intel. PDF2MD addresses these through a three step pipeline: it first strips watermark objects out of the PDF file itself, then filters out repeated headers and footers using frequency analysis, and finally extracts text, tables, and images using a combination of the pdfplumber and PyMuPDF libraries plus some custom rules. For tables specifically, it uses two different detection strategies depending on the document's layout style, one for tables where each row is a single text block split by line breaks, and another for tables where columns are separated by wide gaps of whitespace across several blocks. Extracted images larger than 3 kilobytes are saved to a matching image subfolder and referenced from the Markdown file, and the output folder structure mirrors the input structure exactly. To use it, you install three Python packages, place your PDF files in an input folder, and run two scripts in sequence, one for watermark removal and one for the full conversion. The author lists clear limitations: encrypted PDFs cannot have their watermarks removed, character level watermarks may leave some garbled text behind, tables that span multiple pages may repeat their header row, vector diagrams are not extracted as images, and scanned PDFs are not supported at all since there is no OCR step. The project is released under the MIT license.