k2pdfopt: optimizes PDFs for viewing on e-readers

[ioo]

Hi,

I am struggling to get k2pdeopt working. I have native pdf files which searchable and selectable text so I would like to use the -n argument.

I attempt to run the program like this:

k2pdfopt -n myfile.pdf

The program shows me that I have selected the options:

Selected options: -n myfile.pdf

I press Enter and I get an error message:

** File or folder -n could not be opened. **

The program does run but it outputs image files which I do not want. I tried combining it with other options. Everything else seems to work but it will not recognize the -n option that I want to use.

I have tried searching for this error message but I wasn't able to find anything.

What am I doing wrong?

I am using version:

k2pdfopt v2.51 (w/DjVuLibre) (c) 2019, GPLv3, http://willus.com
Compiled Jan 7 2019 with Gnu C v8.2.0 for Linux on x64

from my package manager on KDE Neon (Ubuntu)

[willus]

Originally Posted by ioo

I am struggling to get k2pdeopt working..
I am using version:

k2pdfopt v2.51 (w/DjVuLibre) (c) 2019, GPLv3, http://willus.com
Compiled Jan 7 2019 with Gnu C v8.2.0 for Linux on x64

from my package manager on KDE Neon (Ubuntu)

See above--you are using a version compiled without MuPDF. This means the -n option won't work with your version. Did you try a linux binary from my web site? Here's an example of how the banner should look:

k2pdfopt v2.53 (w/MuPDF,DjVuLibre,OCR) (c) 2020, GPLv3, http://willus.com
Compiled Jul 18 2020 with Gnu C (Mingw64) v9.3.1 for Win64 on x64.

[ioo]

Originally Posted by willus

See above--you are using a version compiled without MuPDF. This means the -n option won't work with your version. Did you try a linux binary from my web site? Here's an example of how the banner should look:

k2pdfopt v2.53 (w/MuPDF,DjVuLibre,OCR) (c) 2020, GPLv3, http://willus.com
Compiled Jul 18 2020 with Gnu C (Mingw64) v9.3.1 for Win64 on x64.

Thank you for the quick reply, the error message was confusing me. It seemed to be indicating that it was misinterpreting my -n argument rather than indicating there were parts of the program missing.

I have been working with k2pdtopt for a few days now and getting some mediocre results with converting a magazine containing 2 and 3 column material with mixed results.

My target device is an 8th generation Kindle. If I do not use -n but instead use the following:

-dev k2 -fs 12 -wrap- -bp m -ac 0.009 -col 3 -ws 0.2 -fc-

I am able to achieve good results in terms of cutting up the paragraphs while preserving the paragraph structure and removing the white space.

This approach however generated images which do not look good on my Kindle (the fonts are not smooth) and it take away text searching capabilities.

I did try the -n option but the resulting pdf is extremely slow to respond on the target device. It basically renders the kindle frozen for minutes. Is this due to the processor not being able to handle the formatting that is added to the pdf file? If I use the original pdf it is responsive. Is my analysis here correct- that the kindle can't handle the added formatting or do you think there is another issue?

To address this, I then tried to use calibre to convert the -n pdf to mobi format but after 1 day of processing a single k2pdfopt file I gave up. Are you aware of this approach working for others? It seems that Calibre will process the original pdf but not the one outputted by k2pdfopt

I did notice something strange. If I do not use the -n and generate a pdf with images, then use calibre to convert it to the mobi format, the fonts looks slightly more smooth. I do not understand how this would work as k2pdfopt has already created the images. How can converting a series of images from pdf to mobi improve the fonts?

I thought maybe the issue was I am using the wrong -dev which outputs images of a certain resolution then maybe calibre is correcting it and resizing the images? Perhaps this can account for the improvement in image quality?

I am using -dev 2 which states it is for Kindle 1-5 however I have an 8th generation kindle. Is this correct or should I use a different -dev?

Calibre states that the "Kindle" profile has a resolution of 525x640 however online I see the resolution for my device to be 600X800. If I use -dev 2 what will the output resolution of the images be?

Any advice you can provide or things I can try to get better results would be greatly appreciated. Thank you for this awesome project.

[willus]

Originally Posted by ioo

My target device is an 8th generation Kindle. If I do not use -n but instead use the following:

-dev k2 -fs 12 -wrap- -bp m -ac 0.009 -col 3 -ws 0.2 -fc-

I am able to achieve good results in terms of cutting up the paragraphs while preserving the paragraph structure and removing the white space.

This approach however generated images which do not look good on my Kindle (the fonts are not smooth) and it take away text searching capabilities.

Maybe you can post an example of your source? It should be keeping the text searching by default if you have MuPDF. Not sure what to say about the fonts--again, if you could post some examples of both the source and converted PDF, that would help the most.

Originally Posted by ioo

I did try the -n option but the resulting pdf is extremely slow to respond on the target device. It basically renders the kindle frozen for minutes. Is this due to the processor not being able to handle the formatting that is added to the pdf file? If I use the original pdf it is responsive. Is my analysis here correct- that the kindle can't handle the added formatting or do you think there is another issue?

I presume you found a version with MuPDF so that -n works? In that case, yes, your analysis is probably correct. The -n option can put more strain on the particular PDF reader. It's typically faster to display a bitmapped PDF.

Originally Posted by ioo

To address this, I then tried to use calibre to convert the -n pdf to mobi format but after 1 day of processing a single k2pdfopt file I gave up. Are you aware of this approach working for others? It seems that Calibre will process the original pdf but not the one outputted by k2pdfopt

You might want to read through this page.

Originally Posted by ioo

I did notice something strange. If I do not use the -n and generate a pdf with images, then use calibre to convert it to the mobi format, the fonts looks slightly more smooth. I do not understand how this would work as k2pdfopt has already created the images. How can converting a series of images from pdf to mobi improve the fonts?

I thought maybe the issue was I am using the wrong -dev which outputs images of a certain resolution then maybe calibre is correcting it and resizing the images? Perhaps this can account for the improvement in image quality?

I am using -dev 2 which states it is for Kindle 1-5 however I have an 8th generation kindle. Is this correct or should I use a different -dev?

Calibre states that the "Kindle" profile has a resolution of 525x640 however online I see the resolution for my device to be 600X800. If I use -dev 2 what will the output resolution of the images be?

Using -dev k2 should be about right. It will actually use 560 x 735 pixels because the kindle doesn't necesarily use all of that 600x800 screen real-estate to display the text. You can go to this page and post screen shots from your kindle and I'll recommend to you the best settings for it. If you could post the pre-calibre and post-calibre PDFs, that might help explain things. To see a list of all device dimensions known by k2pdfopt:


k2pdfopt -dev "?"

[ioo]

Hi, sorry I am not having any success using this software to quote you.

"It should be keeping the text searching by default if you have MuPDF."

I downloaded the latest version from your website, it does show MuPDF however if I use the list of arguments I mentioned I will end up with a series of images inside a pdf. When I transfer that pdf to my Kindle I lose the ability to search for text using the Kindle search won't I?

"I presume you found a version with MuPDF so that -n works? In that case, yes, your analysis is probably correct. The -n option can put more strain on the particular PDF reader. It's typically faster to display a bitmapped PDF."

It is a huge difference, like if I try to open one of the pdf files on my Kindle the whole Kindle becomes unresponsive and frozen for 5 or 6 minutes. I can't turn the page or go to the home menu. The device is completely disabled. I don't see this with the original PDF, does this sound right or do you think there is some other problem? It would really be nice to have native text on my e-reader so I can search.

I will work on uploading sample images. I just noticed something when comparing the k2pdfop1 pdf to the k2pdfopt file converted to awz3 with Calibre.

When viewing the k2pdfopt pdf on the Kindle, the text is uneven. Like the letter "I" might have a thicker border on the bottom then the top. The word "Room" will have double thickness on the "o o" letters however the next word "door" will have thin o's. The work "Santa" will have a super thin "S" and double thick "t". A letter "e" will look like it has a fuzzy halo around it.

I can pinch to zoom and the letters render differently at a different magnification level. If I zoom in a lot it actually looks more consistent (although much more pixelated).

When I look at the same page of text converted to awz3, each letter is more consistent. If I really look carefully, I can see each "o" letter is rendered very slightly differently (one or two pixels out of place) but I can just barely see this. Looking casually the text looks quite consistent.

I cannot use the device zoom in this format, whatever I try it shows me a list of text sizes to change to but since I have a series of images it doesn't have any effect.

When I zoom in the pdf format I notice that the device shows me the text then it overlays some sort of processing on top of the letters. It looks slightly "thicker or darker" afterwards.

Could the differences be attributed to the different programs on the device rendering the same images differently?

If I look at the pdf file using the KDE pdf viewer- okular and the converted file using Calibre's built in ebook viewer, the text also looks slightly more consistent on the Calibre file.

Calibre is outputting a file that has slightly different dimensions than k2pdfopt, could this be the difference?

Could this be a resolution issue?

Is it possible to convert these images to a lossless image format or is this not supported by ebook readers?

[willus]

It would be a lot easier to help you if you would post some examples.

[PlayaCat]

I use k2pdfopt and like it a lot. Only problem is that it doesn't always get indents correct - a real PITA if you are trying to read a book with Python coding examples, where improper tab indenting changes the meaning of the code (Why didn't Van Rossum use braces?). I just use Sumatra on a laptop or a 7" or 10" tablet to read those PDF Python books as PDFs and don't convert them for Kindle.

[PlayaCat]

Bitdefender Total Security automatically quarantined k2pdfopt v2.53 and my desktop link to it the first time I ran it from the desktop shorcut link. I had to put it in my exceptions to keep Bitdefender from auto-scanning it. It didn't post a reason for the quarantine. I downloaded it directly from https://willis.com/k2pdfopt so I know I didn't get a rogue malware copy. What's up with that?

[willus]

On the Python conversion, do you use -mode fitwidth ? That will not change any indentation, but it also won't wrap text.

The first times I download new versions of k2pdfopt from my own web site, my antivirus software also tries to quarantine them and I have to tell it not to. Hopefully as more people download it, it will get more trust points with the AV community. Sorry.

[kleber]

Hello! First of all, thanks for this software, it seems to be the solution for reading scanned pdfs on kindle!

But unfortunnely I was not able to use It, because every pdf I try to convert gives me the same error:


Reading 876 pages from C:\Users\Kleber\Desktop\Meus documentos\1468-Curso-de-Direito-Civil-Parte-Geral-e-LINDB-Vol1-2017-Cristiano-Chaves-de-Farias-Nelson-Rosenvald.pdf ...

Detecting document orientation ... No rotation necessary.

Cannot open PDF file C:\Users\Kleber\Desktop\Meus documentos\Torrents\1468-Curso-de-Direito-Civil-Parte-Geral-e-LINDB-Vol1-2017-Cristiano-Chaves-de-Farias-Nelson-Rosenvald_k2opt.pdf for output!



** Conversion of file C:\Users\Kleber\Desktop\Meus documentos\1468-Curso-de-Direito-Civil-Parte-Geral-e-LINDB-Vol1-2017-Cristiano-Chaves-de-Farias-Nelson-Rosenvald.pdf aborted (File overwrite not allowed). **


Total CPU time used: 18.42 s

Any clues? I´m using win 7.

Thanks in advance!

[willus]

Originally Posted by kleber

... every pdf I try to convert gives me the same error:

Reading 876 pages from C:\Users\Kleber\Desktop\Meus documentos\1468-Curso-de-Direito-Civil-Parte-Geral-e-LINDB-Vol1-2017-Cristiano-Chaves-de-Farias-Nelson-Rosenvald.pdf ...

Detecting document orientation ... No rotation necessary.

Cannot open PDF file C:\Users\Kleber\Desktop\Meus documentos\Torrents\1468-Curso-de-Direito-Civil-Parte-Geral-e-LINDB-Vol1-2017-Cristiano-Chaves-de-Farias-Nelson-Rosenvald_k2opt.pdf for output!

Any idea why k2pdfopt is trying to save in the "Torrents" folder? Your source file does not appear to be in that folder so you would have had to explicitly direct k2pdfopt to save into that folder. If the Torrents folder doesn't exist, that is the problem. It's either that or (1) you don't have permission to write into that folder, or (2) you are out of disk space (less likely since it can't even open the file).

[kleber]

Originally Posted by willus

Any idea why k2pdfopt is trying to save in the "Torrents" folder? Your source file does not appear to be in that folder so you would have had to explicitly direct k2pdfopt to save into that folder. If the Torrents folder doesn't exist, that is the problem. It's either that or (1) you don't have permission to write into that folder, or (2) you are out of disk space (less likely since it can't even open the file).

Thanks for the answer!

Aparentelly my antivirus was blocking k2pdfopt from opening the file. But now I´m doing some tests and I wonder: is there a way to do an autocrop of a scanned pdf to take away the margins (including those dark edges due to copying artifacts) while, at the same time, maintaining the content of each original page (not mixing pages together)? Thanks!

[willus]

Originally Posted by kleber

Thanks for the answer!

Aparentelly my antivirus was blocking k2pdfopt from opening the file. But now I´m doing some tests and I wonder: is there a way to do an autocrop of a scanned pdf to take away the margins (including those dark edges due to copying artifacts) while, at the same time, maintaining the content of each original page (not mixing pages together)? Thanks!

Yes--see the -ac option in the command-line usage page. Something like this:

k2pdfopt -mode tm -ac myfile.pdf

The autocrop (-ac) option can give mixed results. You'll have to see if it works for you. If the page contents are in a consistent location page to page, you can also assign a manual crop box using the section of the gui shown in the attachment.

[ipparcos]

Hi! I want to insert the page numbers (below centered) in the final pdf. How to do? Thanks.

[willus]

Originally Posted by ipparcos

Hi! I want to insert the page numbers (below centered) in the final pdf. How to do? Thanks.

There is no option to do that with k2pdfopt. You’ll have to use some other software to add page numbers.