k2pdfopt: optimizes PDFs for viewing on e-readers

[Masacroso]

Originally Posted by willus

The PDF file handling routines support UTF-8 strings, whereas the DJVU routines do not (yet). So this is consistent. Thank you for checking.

No, thank to you willus for your time and effort making this amazing program.

[kleinjar]

When I do this:

k2pdfopt -dev kv -mode 2col -c -fc- -fs 7.0

to a certain 2 column pdf, a portion of the source's first page does not get split into two columns. Images attached.

No biggie to me but I thought I'd report it.

Source document



BTW thanks for implementing fixed-size text output Willus.

[willus]

Originally Posted by kleinjar

When I do this:

k2pdfopt -dev kv -mode 2col -c -fc- -fs 7.0

to a certain 2 column pdf, a portion of the source's first page does not get split into two columns. Images attached.

No biggie to me but I thought I'd report it...

The page number on the first page is causing the issue, even though I don't think it should be. I will look into it further when I get a chance, but for now you can crop it out by adding the crop-box directive below to your command line. Thanks for reporting this.

-cbox1 0.44in,0.36in,7.2in,9.4in

[krymeljana]

Dear fellow lovers of k2pdfopt,

I've recently come to know this awesome software to convert scientific papers into a format which can be comfortably read on the Kindle.
Thank you so so much for that!!

Now I was trying to increase the font size by a bit but I need help with that.
I tried it out with my Kindle-Paperwhite2 and several different pdfs and which ever settings I use except the preset "2-colum paper" configuration it produces weird results.

How do I adopt the "2-column paper" settings to increase the output font size?

I would not mind line breaks to be in a different place and also not if images were wider than the kindle screen so that you have to swipe while viewing images.

What I tried:

• Using Paperwhite 2 output (produced a similar looking pdf as "Kindle 1-5" but on the Kindle it showed a tiny pdf-page in the middle of the screen).
• Using "fixed output font-size", both with "native pdf" enabled and disabled (Produced a similar looking pdf- same font size also- but on the Kindle it again showed just a tiny pdf-page in the middle of the screen)
• Increasing the "DPI" from 167 to 267, both with "native pdf" enabled and disabled (With "native pdf" enabled the results were the same as above; "native pdf" disabled gave me a strangely shifted text in the pdf)

I am out of ideas. Can anyone help me please?

Best,
Jana

[willus]

Originally Posted by krymeljana

...
How do I adopt the "2-column paper" settings to increase the output font size?
...

Welcome to MR. Can you post or PM (private message) to me one of the PDFs you are having trouble with? If the columns are narrow and do not require re-flow at the font size you desire, you should be able to use the output DPI to scale the font to the size you want, but you'll want to add this option:

-fc-

(You can put that in the "Additional Options" box in the MS Win GUI.) That will prevent the columns from being fit to the width of your device automatically, which is the default behavior (ignoring the -odpi setting). Sorry about that--it's a tricky thing to remember and can hang people up (even me). I'll have to think about how to make it more obvious.

If the columns are wide enough that the text needs to be re-flowed to get the font size you want on your kindle, then you cannot use native mode, since it is incompatible with text re-flow. So you can't use "-mode 2col" or "-mode fw", both of which turn on native output.

[hmijail]

Hello,

I am trying to use k2pdfopt just to make searchable a scanned PDF. According to the help pages, I should be able to basically use -mode copy -ocr; but it's not working: the resulting PDF contains no OCR'd text.

The best I have managed is to use -as -ac -ocr -p2 , which at least gets *some* of the text in one of the pages, but the result is a pretty scrambled PDF. The text dump itself is flowed to short lines.
If I add the -mode copy at the beginning, no text comes out.

So I would like to ask for some guidance for what to try next. I'm not posting publicly the PDF because it's a scan of personal documentation, but I can send it somewhere if it can help fix something.

Maybe the problem is with the scan itself. So I would suggest: maybe you could post some examples of full OCR scans in your web page together with the command lines that created them? That way one could quickly get an idea of what should be working or not.

I guess that this usage of k2pdfopt is almost off-topic for mobileread.com, so if I should ask somewhere else, please let me know

[willus]

Originally Posted by hmijail

I am trying to use k2pdfopt just to make searchable a scanned PDF. According to the help pages, I should be able to basically use -mode copy -ocr; but it's not working: the resulting PDF contains no OCR'd text.

The best I have managed is to use -as -ac -ocr -p2 , which at least gets *some* of the text in one of the pages, but the result is a pretty scrambled PDF. The text dump itself is flowed to short lines.
If I add the -mode copy at the beginning, no text comes out.

...

Maybe the problem is with the scan itself...

Welcome to MR. Yes, it does sound as if the issue is with your scan if you have to use -as and -ac (-p2 is not a correct option unless you have a space between the 'p' and the '2'). You do have Tesseract installed correctly, I take it? Can you PM me a link to your source PDF and I'll have a look?

I do have an OCR help page, though it doesn't have a lot of varying source formats--maybe I'll start an examples page with something mimicking your source file as the first example.

[vgreader]

hi willus and thank you for the useful k2pdf
I have a problem. in a scanned pdf full of images and with texts in between them sometimes test is not detected. my devise is voyage and settings:
-ui- -x -wrap+ -c -p 138-139 -j 0+ -odpi 300 -w 1016 -h 1364
I am using wallauers GUI.
http://www.4shared.com/office/UEHjxU...r_convert.html

[willus]

Originally Posted by vgreader

hi willus and thank you for the useful k2pdf
I have a problem. in a scanned pdf full of images and with texts in between them sometimes test is not detected. my devise is voyage and settings:
-ui- -x -wrap+ -c -p 138-139 -j 0+ -odpi 300 -w 1016 -h 1364
I am using wallauers GUI.
http://www.4shared.com/office/UEHjxU...r_convert.html

Welcome to MR. I've sent you a private message. Please read.

[vgreader]

-ui- -x -wrap+ -c -j 0+ -cgr 0.6 -ch 1.5 -odpi 300 -w 1016 -h 1364

HI Willus.
I solved the problem usung this settings!
Thank you

[willus]

Originally Posted by vgreader

-ui- -x -wrap+ -c -j 0+ -cgr 0.6 -ch 1.5 -odpi 300 -w 1016 -h 1364

HI Willus.
I solved the problem usung this settings!
Thank you

Wow, that makes the text very large, but okay. Good work! BTW, the default for -ch is 1.5 already, so you don't need to specify -ch 1.5.

[moll]

Hi,

I want to convert two columns PDF into one column while keeping everything else (color, etc.).

The idea is just cut each PDF page into two vertically, then concatenate them one by one.

How should I do that with k2opt?

[willus]

Originally Posted by moll

Hi,

I want to convert two columns PDF into one column while keeping everything else (color, etc.).

The idea is just cut each PDF page into two vertically, then concatenate them one by one.

How should I do that with k2opt?

Welcome to MR. Sorry I've changed my answer a couple times. This is the best way:

k2pdfopt -mode crop -grid 2x1x0 mysourcefile.pdf

If you are working with the MS Windows GUI, you can select "crop" from the conversion mode, and then put the -grid 2x1x0 in the "Additional Options" box.

[drjd]

Originally Posted by willus

... This is the best way:

k2pdfopt -mode crop -grid 2x1x0 mysourcefile.pdf

If you are working with the MS Windows GUI, you can select "crop" from the conversion mode, and then put the -grid 2x1x0 in the "Additional Options" box.

Great! That helped me too! Thanks, willus.

[dhdurgee]

I was encountering cases where scanned pages of a magazine with two columns were not being detected as such. On a hunch I added "-as" to my command line and it appears to have fixed the problem. Watching as the program ran I don't recall seeing a rotation exceeding +/-1 degree, but it appears that the columns were tight enough that even a minor rotation was a problem with recognition.

Perhaps this will save someone having similar problems some time fixing it.

Dave