Thanks for your perspective. The project is a closed source proprietary one. What are the options to obtain a paid license for the code you refer to?

The code I am referring to is from Scribe OCR, which I also maintain. Scribe OCR uses Tesseract.js, however provides advanced features such as PDF support. Among other things, this includes the ability to run OCR on .pdf files, and create .pdf output that combines the OCR overlay with the source file (rather than using a series of rendered images). Some of the PDF-related features rely on code from mupdf, which is a dual licensed AGPL/commercial library, so our use of that code requires the free version of Scribe OCR to be AGPL, with proprietary use requiring a commercial license.

While the PDF functionality can be used in an ad-hoc manner (including embedding in other applications), the easiest way to test is to use the web interface at scribeocr.com. Using this interface, you can upload a .pdf document, run OCR, and download a .pdf file where the OCR text is combined with the input document (rather than rendering everything to images). This results in files that are significantly smaller than creating a new PDF from images, and generally maintains the appearance of the input .pdf much better.

Please try with some test documents, and let me know if this is what you had in mind. If that looks promising for your use case, we can discuss licensing.