ocr-and-documents

Use pymupdf to open the PDF and call page.get_text() on each page, or run the extract_pymupdf.py helper script for plain text, Markdown, tables, or metadata output. It installs in seconds and needs no models.

pymupdf is a ~25MB library that handles text-based PDFs instantly, including tables and embedded images. marker-pdf is a 3-5GB PyTorch pipeline that adds OCR for scanned documents, equations, forms, and reading-order detection.

Use marker-pdf, which performs OCR in over 90 languages and outputs Markdown. Run extract_marker.py on the scanned file, optionally with --use_llm for higher accuracy, after confirming about 5GB of free disk space.

No, pymupdf cannot OCR images, so scanned PDFs return empty or minimal text. For those files you need marker-pdf, which runs OCR over page images and reconstructs the layout.

marker-pdf requires roughly 5GB of free disk for PyTorch plus about 2.5GB of models downloaded to the HuggingFace cache on first use. The helper script's --check flag verifies disk space before installing.

Yes, try web_extract on the URL first, which converts PDFs to Markdown via Firecrawl with no local dependencies. Local extraction with pymupdf or marker-pdf is only needed for local files, failures, or batch jobs.