Adding OCR text to a PDF (even a noisy one)

pdf-ocr.sh

#!/usr/bin/env bash
# usage: ./pdf-ocr.sh scanned.pdf
#
# Download this file
# Make it executable:
#   chmod +x pdf-ocr.sh
# Run it on your example file
#   ./pdf-ocr.sh scanned.pdf
set -euf -o pipefail -o xtrace

INFILE=$1
BASENAME=$(basename "$1" .pdf)
TIFFFILE=$BASENAME.tiff
OCRDPDFNOEXT=$BASENAME-OCRd-big
OCRDPDF=$OCRDPDFNOEXT.pdf
SMALLEROCRDPDF=$BASENAME-OCRd.pdf

# Make a multipage TIFF of the original PDF ~700MB
gs -o "$TIFFFILE" -sDEVICE=tiff32nc -r300 "$INFILE"
# OCR the TIFF using tesseract4
tesseract "$PWD/$TIFFFILE" "$PWD/$OCRDPDFNOEXT" pdf
rm "$TIFFFILE"
# Convert images in PDF to jpeg to reduce size ~4MB
gs -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 -dPDFSETTINGS=/ebook -sOutputFile="$SMALLEROCRDPDF" "$OCRDPDF"