pdf-to-text

PDF to Text

Extract text from PDF files. Try the fast path first, fall back to OCR for scanned documents.

Strategy

1. pdftotext (fast, text-based PDFs)

pdftotext "$INPUT" -

For layout-sensitive documents (tables, columns):

pdftotext -layout "$INPUT" -

2. OCR fallback (scanned documents)

If pdftotext returns empty or garbage, the PDF is scanned images. Use ocrmypdf to add a text layer, then extract:

OCR_OUT=$(mktemp /tmp/ocr-XXXXXX.pdf)
ocrmypdf --skip-text "$INPUT" "$OCR_OUT"
pdftotext "$OCR_OUT" -
rm "$OCR_OUT"

Or OCR individual pages directly:

PAGE_DIR=$(mktemp -d /tmp/pages-XXXXXX)
pdftoppm "$INPUT" "$PAGE_DIR/page" -png
for f in "$PAGE_DIR"/page-*.png; do tesseract "$f" stdout; done
rm -r "$PAGE_DIR"

3. Specific pages

Extract a page range before processing:

SUBSET=$(mktemp /tmp/subset-XXXXXX.pdf)
qpdf "$INPUT" --pages . 5-10 -- "$SUBSET"
pdftotext "$SUBSET" -
rm "$SUBSET"

How to tell if OCR is needed

pdftotext "$INPUT" - | wc -w

If the word count is near zero for a multi-page document, it's scanned.

When to use which

Situation	Strategy
Normal PDF, digital text	pdftotext
Scanned document, forms	OCR (ocrmypdf or tesseract)
Large PDF, only need some pages	Extract pages first with qpdf
PDF with complex tables	pdftotext -layout, or tabula-java

Requirements

• pdftotext (from poppler)
• tesseract (for OCR)
• ocrmypdf (optional, wraps tesseract)
• qpdf (for page extraction)

macOS: brew install poppler tesseract qpdf and optionally pipx install ocrmypdf

Debian/Ubuntu: sudo apt install poppler-utils tesseract-ocr qpdf and optionally pipx install ocrmypdf

Red Flags

• PDF is encrypted or password-protected — decrypt first with qpdf --decrypt
• Output is garbled unicode — PDF uses custom fonts without encoding. Try OCR instead.
• Tables come out jumbled — pdftotext doesn't understand table structure. Use -layout or a dedicated table extractor.