k2pdfopt: optimizes PDFs for viewing on e-readers

[crankypants]

Google books often has PDF files which are just a set of images of scans from an old book. Does this software convert those scanned images (inside the PDF) to text or EPUB? Calibre does this but only with 98% accuracy and Calibre doesn't support ligatures (like "if" next to each other which then becomes one electronic character). So if I have 1,000,000 words total in the book, then I have to find and correct 20,000 words that didn't get identified correctly. And that usually means going back to the PDF to read the actual text and type it in.

[willus]

Originally Posted by crankypants

Google books often has PDF files which are just a set of images of scans from an old book. Does this software convert those scanned images (inside the PDF) to text or EPUB? Calibre does this but only with 98% accuracy and Calibre doesn't support ligatures (like "if" next to each other which then becomes one electronic character). ...

K2pdfopt generates PDF output only, but it will add an OCR layer to the scanned text, or you can output the OCR'd text directly to an ASCII text file. It uses the Tesseract OCR engine, so that will govern its accuracy. I don't know if it is better than calibre--I'm not sure which OCR engine calibre uses. I'm also not sure what you mean by "supports ligatures." Do you mean you want it to generate a special "ligature" character code, or you want it to correctly break ligatures into their two separate letters? To be honest, I don't recall Tesseract's behavior on ligatures at the moment, either way. It's easy enough to try it out.

PS. Are you sure calibre is doing the OCR and the OCR layer isn't already in the scanned file? As far as I can tell, calibre does not have integrated OCR capability unless you are using it with a third-party tool. If the OCR is in the scanned file, it's probably done with Tesseract already, since Tesseract is supported by Google.

[oren64]

I want to say thanks for this excellent tool, now i can read books as pdf in my Kobo in much more pleasant way.

The converting of pdf in Hebrew RtL is wonderful, before i convert the pdf to epub and the result is not satisfying, and the process is exhausting pdf>word>pdf>word>htm>epub (the twice pdf>word is because the text is showing inverted at first).

My request is to add an option, to add cover as image file.
part of my pdf files don't have a cover, so i convert the image file to pdf with software and merge image-pdf to the book with Simpo PDF Merge.

[willus]

Originally Posted by oren64

I want to say thanks for this excellent tool ...
My request is to add an option, to add cover as img file.
part of my pdf files don't have a cover, so i convert the img file to pdf whit software and merge image-pdf to the book with Simpo PDF Merge.

Thank you. That's a good idea. I'll add it to my feature request list.

[kdalloway]

crash on win10，older cpu version can not download

[willus]

Originally Posted by kdalloway

crash on win10，older cpu version can not download

Can you possibly post or PM me the source PDF file and the options you used to convert it? if your PC was made after 2008, the old CPU version probably won't help. 32 or 64 bit version?
Edit: I just verified the older-CPU version. It downloaded on the first try and runs correctly.

[eschwartz]

Originally Posted by willus

K2pdfopt generates PDF output only, but it will add an OCR layer to the scanned text, or you can output the OCR'd text directly to an ASCII text file. It uses the Tesseract OCR engine, so that will govern its accuracy. I don't know if it is better than calibre--I'm not sure which OCR engine calibre uses. I'm also not sure what you mean by "supports ligatures." Do you mean you want it to generate a special "ligature" character code, or you want it to correctly break ligatures into their two separate letters? To be honest, I don't recall Tesseract's behavior on ligatures at the moment, either way. It's easy enough to try it out.

PS. Are you sure calibre is doing the OCR and the OCR layer isn't already in the scanned file? As far as I can tell, calibre does not have integrated OCR capability unless you are using it with a third-party tool. If the OCR is in the scanned file, it's probably done with Tesseract already, since Tesseract is supported by Google.

Yup -- calibre will rely on existing OCR in the file, but otherwise simply adds the images themselves.

[isaacbh]

I'm using the cbox option quite a bit on complex PDFs. I was wondering why k2pdfopt will ignore any non-cbox'd pages. Is there a reason for this design? Right now I have to use -cbox 0,0 for any other page or page range. Wouldn't it be easier if k2pdfopt will assume that all pages without an explicit cbox should be treated with an implicit cbox 0,0? (This applies to ibox as well).

[willus]

Originally Posted by isaacbh

I'm using the cbox option quite a bit on complex PDFs. I was wondering why k2pdfopt will ignore any non-cbox'd pages. Is there a reason for this design? Right now I have to use -cbox 0,0 for any other page or page range. Wouldn't it be easier if k2pdfopt will assume that all pages without an explicit cbox should be treated with an implicit cbox 0,0? (This applies to ibox as well).

I'll consider adding an option to default to -cbox 0,0 on all pages that don't already have a -cbox specified. But I don't think I want -ibox to default to anything in particular for all unspecified pages.

[isaacbh]

Oh right, defaulting ibox to 0,0 is kinda stupid But thanks for considering it for cbox

[kahnwong]

Hi. Crop function works wonder. However, I have some problem with re-flow text. I converted a PDF using re-flow (-dev kv -wrap+) and when I read it on Paperwhite 3, I got this. Wonder whether it's on my end or not.

[willus]

Originally Posted by kahnwong

Hi. Crop function works wonder. However, I have some problem with re-flow text. I converted a PDF using re-flow (-dev kv -wrap+) and when I read it on Paperwhite 3, I got this. Wonder whether it's on my end or not.

That's difficult to diagnose without more information. When you view the PDF on a PC reader, is it also cropped like that? Are you using any other options? Does the conversion look right in the preview window (if you're using Windows)? Can you post (or PM me) the source and converted PDF files?

[fullspeedin2sun]

I have a large document with tiny text that is tough to read and was hoping that I could make the text larger with k2pdfopt. Can I change the text size of the pdf when using native PDF output? The size of the text in the preview doesn't seem to alter whether I set the DPI to 100 or the 300 my Paperwhite 3 runs at.

[willus]

Originally Posted by fullspeedin2sun

I have a large document with tiny text that is tough to read and was hoping that I could make the text larger with k2pdfopt. Can I change the text size of the pdf when using native PDF output? The size of the text in the preview doesn't seem to alter whether I set the DPI to 100 or the 300 my Paperwhite 3 runs at.

Yes, you should be able to magnify the text with k2pdfopt, but it depends on what settings you are using and how the document is laid out. If magnifying the text means you also have to re-flow it, then native mode won't work.

Can you post the source PDF or a couple of pages from it, or PM it to me? Also can you send a screen shot of the options you are using?

[fullspeedin2sun]

The options I'm using are in the attached screencap. I was hoping to make this PDF easier to read on my Kindle Paperwhite 3 without making it much bigger as it looks like it was natively created.

Also, is there an option to add margins in the GUI?