OCRmyPDF

Install OCRmyPDF and Tesseract to create searchable PDFs from scanned documents.

ansible-playbook -i inventory -K playbooks/lab-stack.yml --tags ocr

🔧 Manual setup and configuration

  • Log out and log back in after running the role so Nautilus/Nemo reload user actions.
  • In Nautilus, right-click a PDF and use OCR PDF (ocrmypdf).
  • The action runs ~/.local/bin/ocrmypdf-wrapper and writes *_ocr.pdf next to the original file.
  • The wrapper defaults to German + English OCR (-l deu+eng) and skips pages that already contain text.
  • Review ~/.cache/ocrmypdf-wrapper.log if conversion fails.

✅ Check

ocrmypdf --version
tesseract --version

Expected: both commands print a version.

echo "test" > /tmp/ocr-test.txt
enscript /tmp/ocr-test.txt -o - | ps2pdf - /tmp/ocr-test.pdf
ocrmypdf -l eng /tmp/ocr-test.pdf /tmp/ocr-test_ocr.pdf
pdfinfo /tmp/ocr-test_ocr.pdf | sed -n '1,5p'

Expected: /tmp/ocr-test_ocr.pdf is created successfully.

ls -l ~/.local/share/file-manager/actions/ocrmypdf.desktop
ls -l ~/.local/share/nemo/actions/ocrmypdf.nemo_action

Expected: at least one action file exists (depending on your file manager).

Nautilus right-click action behavior

  • Menu entry: OCR PDF (ocrmypdf)
  • File type: appears for PDF files (application/pdf)
  • Execution: calls ~/.local/bin/ocrmypdf-wrapper "%f"
  • Output naming: input paper.pdf becomes paper_ocr.pdf
  • Batch behavior: the wrapper accepts multiple files and processes each one in sequence
  • Troubleshooting: if the menu entry does not appear, restart Nautilus (nautilus -q) or relogin

Why this software?

OCRmyPDF is preferred for the workstation setup because it runs locally, adds a searchable text layer while preserving the original PDF, and provides a mature command-line workflow for rotation, deskewing, optimization, and PDF/A output. It also works without an account, API key, document upload, or separate model service and integrates directly with the automated file-manager action above.

Consider an alternative when the task is broader than making scanned PDFs searchable:

  • Mistral OCR can extract document structure and content for downstream AI workflows. It is a hosted API, so review costs, credentials, and data-handling requirements before uploading research or institutional documents.
  • PaddleOCR provides trainable, deep-learning OCR and document-layout models for many languages and complex extraction pipelines. It requires more model/runtime setup, and producing a searchable archival PDF needs additional pipeline work.
  • Baidu Unlimited OCR is a managed OCR option for high-volume or application-integrated extraction. It requires a Baidu cloud account and sending documents to an external service; confirm current service limits, pricing, supported regions, and privacy terms before use.