k2pdfopt: optimizes PDFs for viewing on e-readers

[nmyshkin]

Originally Posted by willus

This CPU came out in 2013. It has a Geekbench 6 score of 170 (single threaded). Kudos to you for getting that much mileage out of your computer but wow that is a low score. Modern CPUs are 20x faster single threaded and 100x faster multithreaded. You might really enjoy a newer system.

It will have to happen someday, but the thought of decrapifying yet another bloatware multimedia extravaganza until it's actually a functional computer again is not that appealing at my age. Speed would be nice when working on a dictionary database or re-encoding a video file, I admit. But speed is not everything.

[willus]

Amen to the decrapifying, though I don't think it will be quite as bad as you might think. Just uninstall, uninstall, uninstall.

[dhdurgee]

I have used k2pdfopt to convert a PDF file for use on my PW4. I used the -ocr option with it as the original PDF did not include any text for searching.

I am now wondering if it would be possible to extract the ocr text from the k2pdfopt output file and use it as the starting point to create a text version of the file to create an azw3 file.

Is this possible with k2pdfopt or another pdf tool?

Dave

[willus]

Originally Posted by dhdurgee

I am now wondering if it would be possible to extract the ocr text from the k2pdfopt output file and use it as the starting point to create a text version of the file to create an azw3 file.

Is this possible with k2pdfopt or another pdf tool?

You can extract the OCR text in UTF-8 format using the -ocrout <file> option. See the command-line usage. You might take a look at my PDF conversion tips page, though it's a bit stale.

[dhdurgee]

Originally Posted by willus

You can extract the OCR text in UTF-8 format using the -ocrout <file> option. See the command-line usage. You might take a look at my PDF conversion tips page, though it's a bit stale.

I added the -ocrout option and it worked as expected.

Thank you for your assistance.

Dave

[stillstill]

Hello.

I have faced this weird issue first time. If there is a thin horizontal line (it is located just before the footnotes begin, I cannot eliminate it since its position varies on pages) the text that is above that line is rendered into so microscopic font, that is comparable with the line width... Is there a way to remediate from this?... Moreover text alignment switches from justified to left for the entire chapter where there at the end that horizontal line is.

Any idea highly appreciated.

Attaching screenshots and pdf to illustrate this.

[willus]

Originally Posted by stillstill

If there is a thin horizontal line (it is located just before the footnotes begin, I cannot eliminate it since its position varies on pages) the text that is above that line is rendered into so microscopic font, that is comparable with the line width... Is there a way to remediate from this?... Moreover text alignment switches from justified to left for the entire chapter where there at the end that horizontal line is.

Any idea highly appreciated.

Thank you for including screen shots and especially for including a source PDF. That makes it much easier to suggest things. There are a couple options.

1. Erase the line using the -ehl option: k2pdfopt -ehl 1 ... source.pdf

2. Set a minimum text row height using the -rhmin option: k2pdfopt -rhmin 1 ... source.pdf

Either one effectively removes the horizontal line from the output. BTW I used -ml 0.2 -mr 0.2 to get rid of the < > marks on the source pages, as you probably did also.

[stillstill]

Thank you for both methods. Both of them perfectly eliminated horizontal line and likely in this way helped to workaround the small font issue (the line presence alone in the output was not an issue). The sudden alignment change still remained. Text alignment changed from "justified" to "left aligned" for the entire source page where there at the end is that horizontal line. Sometimes it affects not the entire source page but the chapter on the page before horizontal line. This remediated effectively by "-j +" which worked flawlessly.

Thank you so much for your really brilliant Application development and such a dedication to the support.

Originally Posted by willus

Thank you for including screen shots and especially for including a source PDF. That makes it much easier to suggest things. There are a couple options.

1. Erase the line using the -ehl option: k2pdfopt -ehl 1 ... source.pdf

2. Set a minimum text row height using the -rhmin option: k2pdfopt -rhmin 1 ... source.pdf

Either one effectively removes the horizontal line from the output. BTW I used -ml 0.2 -mr 0.2 to get rid of the < > marks on the source pages, as you probably did also.

[dhdurgee]

After I got things sorted out with your tool I had yet another thought occur to me. I passed the original scanned image document along with some guidance to the Grok AI engine and asked it to do the OCR for me detecting the images and inserting placeholders for them in the output. This was after trying to get it to do a full conversion to an epub format, which it turned out was beyond its capabilities. It was, however, able to give me the files comprising the epub piecemeal so that I could assemble the epub manually.

I hit some issues with limitations of grok.com and had to do the processing in pieces, but I now have a final epub of that document with images that is about 1/3 the size of your optimized output.

I understand that the Grok API is available to developers and can tell you that its OCR capabilities were great. It might be worth your time to investigate if it could be useful to you in a later release of your tool.

Dave

[willus]

Originally Posted by dhdurgee

I understand that the Grok API is available to developers and can tell you that its OCR capabilities were great. It might be worth your time to investigate if it could be useful to you in a later release of your tool.

That is interesting information. Thanks. I may look into it at some point, but I'm a bit of a dinosaur when it comes to AI.

[dhdurgee]

Originally Posted by willus

That is interesting information. Thanks. I may look into it at some point, but I'm a bit of a dinosaur when it comes to AI.

I'm new to it as well, but it is simple enough to go to grok.com and ask about something. I have also had it provide me with recipes meeting certain requirements or with particular ingredients to use up left overs.

Its even simpler to use than a search engine and produces more understandable results in most cases.

Give it a try and see.

Dave

[DNSB]

My experience with AI and recipes has not been that great. I was curious as to what it would give me for a couple of recipes. I couldn't bring myself to try a "mild" curry recipe that includes 4 tablespoons of 7 Pod Jonah pepper powder. One of the other recipes it supplied was basically a quote of Guy Fieri's Cheddar Trans-Porter Soup recipe.

I worked in IT in an education setting for years. One of the teachers whom I still chat with is still being surprised by the number of students whose idea of writing a program is to ask an AI to write it for them and then never bother to try using the program. He has a form now that basically says "The AI's first error occurred at line ___ in the source code and since you used AI to write the program, your mark is 0. A second occurrence will result in your final mark for the course being 0".