I received my Kindle 3 (a.k.a. the Kindle Keyboard) almost ten years ago, and it has proven quite durable (even though some parts of the case are now cracked and it's been stained by various liquids...). Since then, I've used it many times to read old books in their original format. What I mean by old books are really classics, that are now commonly found in the online libraries of the world (my favourite is of course Gallica, but there are many others, e.g. Europeana, Deutsche Nationalbibliothek, archive.org for books and texts).

One of the recurring problems I have is to convert the PDFs from these online libraries to a format usable on my Kindle. Searching for solutions has taken me to the scripts and tips I'm using below. I'll try to make a case study using the origin of the species found here. I'll call this file darwin.pdf and will try to convert it to various formats, comparing file sizes. The file we start with is a 145 MB PDF with color scans.

Ghostscript

The main tool I will use is Ghostscript, a free "interpreter for the PostScript language and for PDF". In my test below, I'll use the latest available version on MacOS: Ghostscript 9.51.

I found a lot of conflicting information about the use of Ghostscript online. I now realize this is probably due to the fact that PDF is a really complicated file format. And that there is a lot going on between the actual raster images embedded before they get rendered to the screen, which explains why there isn't an option to "just compress the images more".

First script: using /ebook and converting to grayscale

A first script to test the conversion is to use the built-in PDF settings designed for ebook use.

We can use it like this:

gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -o darwin1.pdf darwin.pdf

Second script: adding grayscale conversion

However, my Kindle can only use grayscale, so let's add a flag for grayscale.

gs -sDEVICE=pdfwrite -dPDFSETTINGS=/ebook -dColorConversionStrategy=/Gray -dProcessColorModel=/DeviceGray -o darwin2.pdf darwin.pdf

Third script: manually setting the DPI

A last option is to manually set the DPI you want so that the image data gets converted.

gs -sDEVICE=pdfwrite -dColorConversionStrategy=/Gray -dProcessColorModel=/DeviceGray -dDownsampleGrayImages=true -dGrayImageResolution=72 -dGrayImageDownsampleThreshold=1.0 -o darwin3.pdf darwin.pdf 

Comparison

Here's a table with the obtained file sizes and usability.

So I have to say that these conversion scripts are quite good. But still, the grayscale image is not very high contrast so it would be great to have a thresholded image.

Other tricks

In the course of the conversion process, I've also found that if you want to crop the images you can use the native MacOS Preview app as described here: http://hints.macworld.com/article.php?story=200711012305556.

Useful links

• A discussion of Ghostscript options for resizing: https://gist.github.com/firstdoit/6390547
• A Python app that shrinks images in an EPUB: https://github.com/murrple-1/epub-shrink (linked by https://ebooks.stackexchange.com/questions/6772/how-can-i-reduce-the-resolution-of-images-in-an-epub-or-mobi-file)

Conclusion

This is a first step for conversion. However, I'm still missing a binarization step where the books get converted to 4 bit grayscale. Maybe this would be possible using the above script epub-shrink process but that would be for another time.