OCR Doesn't seem to make the PDF searchable about stirling-pdf HOT 9 CLOSED

Yep you nailed it ! Mounted the big eng model file only. (not the directory) and it works !

from stirling-pdf.

Sounds like a good fix. Btw when you convert a pdf which you first used ocr to word, the ocr layer disappears and the word document has only an image. Is this intended ?
Also amazing app! You should post it on r/selfhosted. A pdf manipulator is one of the most requested apps I’ve seen over the years. The closest is hrconvert but that doesn’t work very well and it’s ui is dated.

from stirling-pdf.

Weird are you able to give me the pdf to test against or is it private? I can't reproduce this my side

Maybe it's how the bigger eng data works I will test with that as well tonight or tomorrow

from stirling-pdf.

Since I am using this at work, that PDF is private, however it happens with all pdf's that I have tried. Will try the fast model and let you know.

Edit: Same issue with fast model. Any way I can send you debug logs or such ? Just point me in the right direction.

from stirling-pdf.

I think I found issue
Mounting that directory is removing some needed files which are already there and which you are missing
I will change docker file this weekend to ensure those files are kept on mount

from stirling-pdf.

I have a fix to create a temp folder during build and copy everything from temp to final folder on container startup.
It does mean you wont be able to delete of the old files in that folder but you can add any new ones fine now
also renaming eng.traineddata to English-Lite.traineddata

from stirling-pdf.

It's not ideal but the backend for pdf to word document is libre office so I can't change how it handles the OCR layer sadly
you would have to raise a issue with LibreOffice directly to get that fixed
The OCR tech im using in backend has different ways of rendering the text
here
So i will try this usecase and see if i can get it working, will track the issue here https://github.com/Frooodle/Stirling-PDF/issues/118

from stirling-pdf.

Also thanks for the comments! i did post on reddit when i first started this app
https://www.reddit.com/r/selfhosted/comments/10pexhn/new_browserbased_pdf_editor_github_link/
I plan to post again when I release V1.0.0 (Once i finally add PDF cropping, PDF signing and improved PDF image importing)

Feel free to make a post for me though:') haha

from stirling-pdf.

Fixed with extra lang support in latest patch

from stirling-pdf.