k2pdfopt: optimizes PDFs for viewing on e-readers

[Jorram]

Originally Posted by willus

@Jorram -- thank you for the post and nice comments. Unfortunately I have little first hand experience with calibre or MOBI files. But there should be a number of experts here at MR. I wonder if your post wouldn't be better off in the MOBI sub-forum? You could request that a moderator move it if you think so. Hopefully some others will chime in.

Thanks, Willus. I will do that, but first I tried to convert one of the more bothersome PDFs that I have. It's all over the place - pictures, boxes, columns, inconsistent margins, etc.

k2pdfopt did a pretty good job but it's still more unusable than usable so I was wondering if you could offer me some advice. Here are the problems I encountered:

1) The search is absolutely broken, at least in Adobe Reader. I wasn't able to find any phrase correctly - at all times the wrong places are highlighted. Wondering how this would behave on the Kindle itself. No idea how to fix this either.

2) The images are broken down and misplaced. I have attached a screenshot to illustrate this. You'd think it happens because the image is close to the page border but this is not the case - virtually all images are broken down in the same way, regardless of their position.

3) Some formatting is handled badly, in particular columns - my second screenshots shows this effect.

4) The cover (the first page) is gone entirely. I can't figure out why would that be.

I used the native pdf output and mostly kept everything default I think.

Any clues?
Thanks!

[willus]

Originally Posted by Jorram

...k2pdfopt did a pretty good job but it's still more unusable than usable so I was wondering if you could offer me some advice...

@Jorram -- I'll be able to do the best job helping you if you can post the PDF source file or send it to me via a private message.

[willus]

Originally Posted by loudyou

Segfault on Mac OS X 10.9 when using russian tesseract data
Selected options: -dev kpw -ocr t -ocrlang eng+rus

init_cube_objects(true, &tessdata_manager):Error:Assert failed:in file tessedit.cpp, line 237
Segmentation fault: 11 (core dumped)

Turns out this is an issue with k2pdfopt not correctly initializing languages that do not have the "CUBE/COMBINED" type of training files. This has been fixed for the next release.

[taschenbillard]

do i something wrong or is a 2mb pdf supposed to get to a size of 50mb after using k2pdfopt?

[willus]

Originally Posted by taschenbillard

do i something wrong or is a 2mb pdf supposed to get to a size of 50mb after using k2pdfopt?

Please see the k2pdfopt FAQ, third question.

[taschenbillard]

ah i see - native output was what i needed
thank you

works nice, but now i have some times lines with bigger font sizes.
i have v2.18 for win and run it with the presets for pw1, command line version is -dev kpw -gui+ -wrap- -n -ocr- -rt auto+ -col 1 (picture #1)

with -dev kpw -mode fw -gui+ -ls- -rt auto+ there is a additional margin on the left side sometimes (picture #3)

[willus]

Originally Posted by taschenbillard

ah i see - native output was what i needed
thank you

works nice, but now i have some times lines with bigger font sizes.
i have v2.18 for win and run it with the presets for pw1, command line version is -dev kpw -gui+ -wrap- -n -ocr- -rt auto+ -col 1
any tips?

Try using -mode fitwidth, as shown in the second example on this page.

[taschenbillard]

i edited my previous post before i saw your answer

i did run meanwhile with -dev kpw -mode fw -gui+ -ls- -rt auto+ ,there is now a additional margin on the left side sometimes (the font size is ok here)

with -mode fp the margin left and right is always changing from page to page (the font size is ok here too)


and thanks for your help

--edit nevermind i think the book is the problem, i will have to crop odd and even pages differently because of the chapter numbers which appear in the top left/right corners

[willus]

Originally Posted by taschenbillard

... i think the book is the problem, i will have to crop odd and even pages differently because of the chapter numbers which appear in the top left/right corners

If you don't need to see the chapter name on each page, you could just crop off the top xx inches of each page with -mt xx. Then you wouldn't have to crop odd and even pages differently. That large black horizontal bar separating sections might cause problems, though, if it is used on a lot of pages. If you need help cropping odd and even pages differently, search this thread for "-cropbox".

[taschenbillard]

i trimmed all 400 pages semi automatic per hand to remove the page numbers and the recurring chapter names and now it works perfect with
-dev kpw -mode fw -gui+ -ls- -rt auto+
everything in the the same font-size and layout is perfect and the margins are good; looks great, font is very crisp and good to read, actually no difference to mobi / epub format
great tool
also now from 2mb size to 1,5mb

the toc seems no longer to work now, do you know a tool that can edit these?

[willus]

Originally Posted by taschenbillard

the toc seems no longer to work now, do you know a tool that can edit these?

I don't know of a tool. K2pdfopt does not presently preserve inter- or intra-document links/references, but it does preserve the bookmarks/outlines table, but not all readers correctly make use of that. If your reader makes use if it, you can always create your own--see the -toc... options in the command-line usage.

[taschenbillard]

i arrived now at editing a few pages with pagenumbers at the bottom manually and removing the rest with -mt 0.73 , good advice, this goes a bit faster

new toc with -toclist also works great

[MaxStirner]

Hi, I wanted to drop an idea (not a feature request at all) to think over. Do you think it would be beneficial to have some "compression" features in k2pdfopt given the fact that Kindle has a limited space availability (no memory card of course)? Most programs for this purpose are commercial and - worse still - Windows only, but I've been using some of them and the results are quite good, sometimes the PDF size is reduced by about 30% without any noticeable quality loss. Don't know exactly how it works but I've read that what the program does is - among other things - eliminating duplicates in the document structure.
I don't know if this is feasible for you to work on so as I said I'm only dropping an idea. Thaks for the great soft - seriously the best swiss knife for pdfs on small screens.

[willus]

Originally Posted by MaxStirner

Hi, I wanted to drop an idea (not a feature request at all) to think over. Do you think it would be beneficial to have some "compression" features in k2pdfopt given the fact that Kindle has a limited space availability (no memory card of course)? Most programs for this purpose are commercial and - worse still - Windows only, but I've been using some of them and the results are quite good, sometimes the PDF size is reduced by about 30% without any noticeable quality loss. Don't know exactly how it works but I've read that what the program does is - among other things - eliminating duplicates in the document structure.
I don't know if this is feasible for you to work on so as I said I'm only dropping an idea. Thaks for the great soft - seriously the best swiss knife for pdfs on small screens.

@MaxStirner -- thanks for the post. When writing native PDF files from k2pdfopt, I use the MuPDF library, and the native output essentially mirrors the source file, so the output file will be very similar in size to the source file. I'm not sure MuPDF has simple routines to compess the file further--I haven't looked into it carefully, but based on the write flags and the "pdfclean" utility that comes with MuPDF, it doesn't look like it. If there is a MuPDF expert that knows otherwise, please let me know.

The other type of k2pdfopt output is one bitmap per page, and there isn't much point in compressing that further since the images are saved as either PNG or JPEG, both of which are already compressed. There are settings in k2pdfopt that affect the size of the images, though (resolution, number of bit planes, JPEG quality, etc.). See my k2pdfopt FAQ, third question.

[willus]

This revision has several new features and bug fixes. See the web site for details.