ocrmypdf-chromelens-ocr 1.0.4

OCRmyPDF plugin using Google Lens API for OCR

OCRmyPDF-ChromeLens-Ocr

A plugin for OCRmyPDF that uses the Google Lens API to perform Optical Character Recognition (OCR).

Features

• High Accuracy: Leverages Google's advanced Lens models.
• Structure Preservation: Correctly handles multi-column layouts and vertical text flows via strict API parsing and rotation-aware sorting.
• Smart De-hyphenation: Merges words broken across lines while respecting punctuation dashes.

---

Installation

Prerequisites

You must have ocrmypdf installed.

Install from Git

pip install git+https://github.com/atlantos/OCRmyPDF-ChromeLens-Ocr.git

Install from Pip

pip install ocrmypdf-chromelens-ocr

Usage

To use this engine, pass the plugin name to OCRmyPDF. You generally do not need to specify a language, as Google Lens auto-detects it.

ocrmypdf --plugin ocrmypdf_chromelens_ocr input.pdf output.pdf

Configuration Options

You can configure the behavior of the plugin using the following command-line arguments:

Argument	Description	Default
--chromelens-no-dehyphenation	Disables the logic that merges hyphenated words across lines. Useful if you prefer raw output.	Disabled
--chromelens-max-dehyphen-len	The maximum length of word parts allowed for de-hyphenation. If both the prefix (before hyphen) and suffix (after hyphen) are longer than this value, the plugin assumes it is a compound word or dash separator and will not merge them.	10

Example: Disable de-hyphenation

ocrmypdf --plugin ocrmypdf_chromelens_ocr --chromelens-no-dehyphenation input.pdf output.pdf

Example: Stricter de-hyphenation (only merge very short breaks)

ocrmypdf --plugin ocrmypdf_chromelens_ocr --chromelens-max-dehyphen-len 4 input.pdf output.pdf

Credits & Acknowledgements

This project is a Python port and adaptation based on ideas and logic from:

1. chrome-lens-ocr:

   • Provided the critical reverse-engineering of the Google Lens Protobuf API (v1/crupload).
   • Logic for strict layout parsing and request structure.

2. OCRmyPDF-AppleOCR:

   • Provided the architectural inspiration for creating an OCRmyPDF plugin that offloads recognition to an external engine.

Disclaimer

This software is for educational purposes. It uses an undocumented private API from Google.

• Privacy: Your images are uploaded to Google servers. Do not process sensitive/confidential data.
• Stability: The API may change or break at any time without notice.
• Rate Limits: Excessive use may result in your IP being temporarily blocked by Google.

License

MIT License