Is PDF / DOCX support on the roadmap?

[wdoppenberg]

trafficstars

I know this is not trivial since I've been unsuccessful in finding any PDF->image Rust library, but is this something you plan on supporting in the future?

If help is needed please let me know.

[robertknight]

Ocrs could potentially integrate with existing libraries or CLI tools for rendering PDFs somehow. It could also serve as a backend for a project like OCRmyPDF. What use case did you have in mind?

[tomtom215]

I would love PDF support as I need to batch load invoice PDF's and extra the text data which can then be saved as JSON to a DB

[woidda]

@tomtom215 fwiw, you can try to preprocess your pdfs with pdf2image which works pretty well.

[physics515]

If we are looking for sentiment here, I'm also looking for PDF support. I have a mixture of text PDFs and scanned documents in PDF form (aka bitmap PDFs).

I'm ingesting them into embeddings for an AI and I'm planning on trying the pdf-extract crate, and if it fails then fallback to pdf2image piped into this crate.

Hopefully in the future, this could just be done with a single crate.

[robertknight]

Rendering a PDF into an image is a complex task which will either involve using platform-specific libraries or compiling large dependencies. Rather than add that into the ocrs crate I think it would make more sense to build libraries on top that can orchestrate the pipeline. In the Tesseract ecosystem, there is OCRmyPDF which does this.