[Conversion Plugin] KFX Input

[tomsem]

A couple of issues with KFX->PDF (specifically with KFX manga/comic format):

1: 'Binding' document property (what Acrobat calls it) should reflect page reading direction of source (e.g. for manga/japanese it should be right-to-left).

Currently it assumes left-to-right page direction, but that messes with PDF books that are laid out for right-to-left page order, such as manga: 2 page spreads will have left and right pages reversed.

I happen to have Acrobat Pro and am able to change it, but Acrobat Reader apparently will not let you do this (option in Advanced document properties is disabled).

It is sorta odd that PDF viewers do not typically let you change it, or even pay attention to it if it is set (Apple Books does not, PDF Expert does not). In such cases the only recourse seems to be to avoid 2-page layout.

It looks like something you can set with PyPDF in ViewerPreferences dictionary:

https://pypdf.readthedocs.io/en/stab...s.html#example

It would be too much to expect every PDF viewer to pay attention to this, but at least Adobe Reader does. And of course it is only an issue in 2-page layout.

2: This is a bit more involved: there are sometimes images with 2-page spreads, twice the width of non-spread images. Conversion to PDF writes out a single page, which of course does not 'spread' in 2 page layout, and only consumes one page instead of two, so it's complicated to add PDF page labels that correspond to what may be in the image.

The first workaround I did for this was to rotate the spread 90 deg clockwise and insert a blank page before or after the spread (depending on page reading direction) forming a sort of pseudo spread that also preserves expected left/right ness of the 'single' pages and normal page numbering.

But the 'correct' thing is to split the image in half and make two pages instead of just one, which is how I am fixing things up now.

My method is not particularly portable or automate-able: I happen to have a page splitter script for Acrobat Pro, which I apply to extracted spread pages, and then I replace the existing spread pages with the split spread pages (again paying attention to page reading order). Often a spread page is a target of a PDF bookmark so I have to fix that too. But it doesn't take that much time. f

I'm not sure how I'd do it without the page splitter script. Even image split-in-half scripts are typically one image at a time. Kindle Comic Converter probably: it can split 'wide' images that it finds and output to CBZ, and then you could make a PDF out of that. So then KFX->CBZ would be first step.

But 'it would be nice' if From KFX->PDF just handled it.

Note that none of this applies to Print Replica KFX: it does not support 2-page layout (maybe someday, one can only hope).

The other fixed layout format (not sure how you create or publish one -- InDesign and submit FXL ePub probably? Kindle Create doesn't do it.) are KF8 fixed-layout that have text in addition to images and unlike Print Replica, support facing pages and have spreads. I think KindleUnpack coverts these to ePub FXL pretty well, and I think calibre can convert these to PDF pretty well or did last time I tried it. In any case, not anything KFX Input has any role in.

[Jaws]

This is a quirk that I haven't seen mentioned here, and it's a quirk of Kindle for PC (but directly related to importing KFX files). I'm wondering if this is "just something" resulting from programmers not anticipating a workflow or a necessary consequence... and, fortunately, I've found a workaround that works for me if it's an actual glitch/problem.

Kindle for PC (stretching back a lot of versions, but definitely up to 2.3.0) has trouble fully downloading newly-added books during sync if a book was deleted in the Kindle library at the 'zon since the last sync. I run Kindle for PC from a command script (batch file) that does the overwrite-the-update-file-with-trash-to-prevent-automatic-updates dance — but it's not open while I'm in the account at the 'zon, only after I sign out of that account. This results in what appears to be an adequate sync, but trying to read the book (to check editions, for example, which can be a real problem in the 'zon's online interface) in either Kindle for PC or by importing into Calibre through KFX-Input results in a read failure. I get the same behavior if I attempt to open a second Kindle for PC session while the first one is open — and once if I ran Kindle for PC directly (without using that batch file). If I shut down Kindle for PC and look at the "downloaded files" in the folder, the filenames are there, but are 0 bytes in length (sometimes the AZW is a fraction of the successful-download-size of the ebook).

The question: Is anyone else seeing this, or is it an artifact of however I've got my machine and accounts set up, and of my process of logging off the the 'zon (in my browser) at any moment I'm not actually engaging on the 'zon?

The workaround: Close Kindle for PC and restart it. I'm mentioning this because this may not be the only related glitch. Or maybe it's a feature, not a bug.

Possible cause: Some session flag somewhere in Kindle for PC, triggered by the delete-a-book process when not logged in to the 'zon account, that tells the sync function "wait a minute, I'm deleting something asynchronously"... and doesn't get reset upon finishing the deletion, but instead is reset only on Kindle for PC program start. That's nothing that can be managed outside of the 'zon itself, but it may be something to keep an eye on (if it's a correct understanding).

[tomsem]

There is an issue with spreads when authoring KFX with Kindle Create and then converting to ePup using From KFX: the <itemref> for spread pages have no properties attribute, so they get rendered (and squished) onto a single page rather than spread across two (I'm testing with Thorium Reader).

So for example:

<itemref idref="part0067.xhtml"/>

should be

 <itemref idref="part0067.xhtml" properties="rendition:page-spread-center"/>

I have not seen this when converting KFX to ePub with one downloaded by Kindle for Windows; the required property is set. AFAICT there is nothing I can set in Kindle Create to control this (the page is marked as 'single page' instead of 'facing page').

I tried using the same properties value for the cover page, so it is centered, and I think that looks better than having it to one side or the other in 2-page view. It's how Kindle Create previewer shows it, as well as any other pages marked as 'single page'. (I do not know if there is a more semantically correct value to use for this, but at least this one seems to work.)

[jhowell]

Originally Posted by tomsem

1: 'Binding' document property (what Acrobat calls it) should reflect page reading direction of source (e.g. for manga/japanese it should be right-to-left).

That should be easy. I will add that to my to do list.

Originally Posted by tomsem

2: This is a bit more involved: there are sometimes images with 2-page spreads, twice the width of non-spread images. Conversion to PDF writes out a single page, which of course does not 'spread' in 2 page layout, and only consumes one page instead of two, so it's complicated to add PDF page labels that correspond to what may be in the image.

I might be able to handle this simply by adding an option to split images that are in landscape orientation when a comic is in portrait orientation.

It will likely be several weeks before I have time to work on these things.

[jhowell]

Originally Posted by Jaws

Is anyone else seeing this, or is it an artifact of however I've got my machine and accounts set up, and of my process of logging off the the 'zon (in my browser) at any moment I'm not actually engaging on the 'zon?

I have seen Kindle for PC hang after clicking on a book cover in the library view to download a book. It usually works correctly after a retry. This might be related to what you are experiencing.

I do not have any suggestions on getting it to work more reliably in an automated environment.

[jhowell]

Originally Posted by tomsem

There is an issue with spreads when authoring KFX with Kindle Create and then converting to ePup using From KFX: the <itemref> for spread pages have no properties attribute, so they get rendered (and squished) onto a single page rather than spread across two (I'm testing with Thorium Reader).

I will look into this once I have more free time.

[tomsem]

Originally Posted by jhowell

I will look into this once I have more free time.

Thank you.

[risho]

When you convert a kfx manga to cbz using from kfx is this a lossless conversion or are you re-encoding the images? I'm noticing that the images coming from the kfx book are a different size (and naming scheme, jpg instead of jpeg) than what i get from converting the older style azw3 (pre 2023 change books).

[tomsem]

Originally Posted by risho

When you convert a kfx manga to cbz using from kfx is this a lossless conversion or are you re-encoding the images? I'm noticing that the images coming from the kfx book are a different size (and naming scheme, jpg instead of jpeg) than what i get from converting the older style azw3 (pre 2023 change books).

This is from the Alternatives to KFX notes in the first post:

Generally the highest quality graphics that Amazon makes available are delivered in KFX format to recent versions of the iOS/Mac and Android apps. The next best image quality can be obtained with a KF8 (AZW3) format book along with an HD image container file (AZW6) obtained for some books using Kindle for PC or an older version of the Android app. A bit worse is a standalone KF8 file obtained from one of those apps or by using the Download & Transfer function of Amazon's web site. Recent versions of Kindle for PC will deliver books in KFX format with graphics quality somewhere between the highest possible and KF8 equivalent depending on the particular book. Worst of all is the image quality in books downloaded directly by Kindle devices in KFX format, which are typically black and white with low resolution.

I have been doing a lot of this conversion lately, as part of converting comics I've purchased from Amazon to other formats (ePub/PDF/CBZ) for reading on other devices. (I actually prefer ePub or PDF as they can have a ToC).

I greatly prefer KFX (since getting around to installing Kindle for PC) as the conversion to ePub/PDF/CBZ is very reliable and simple compared to trying to do the same from AZW3. (Kindle Classic for Mac does not deliver a file that KFX import can process for any comics or manga -- they have 'kcr' file extension). I have not tried to get older Mac versions which might allow this.)

Generally the images are essentially the same, but for example AZW3 will have 1249 pixels wide where the KFX has 1250. And the images in CBZ-from-KFX are a little larger than those exported by KindleUnpack (94.5 MB vs 83.3 MB in the one I checked), indicating less aggressive compression.

There's no signifcance to the naming of the image files.

[risho]

i appreciate the information about the source quality.

does the KFX Input extension re-encode the images when converting them to cbz or is it a lossless conversion? maybe i missed that on the first page too.

[jhowell]

Originally Posted by risho

i appreciate the information about the source quality.

does the KFX Input extension re-encode the images when converting them to cbz or is it a lossless conversion? maybe i missed that on the first page too.

The original images are left intact whenever possible. One situation where that is not the case is when KFX format contains a grid of images that are supposed to make up a single image in the book. In that case the plugin will combine them and deliver the result as a lossless PNG file so that no further degradation of quality occurs.

Be aware that Amazon does not usually deliver the original images as provided by the publisher. Various reductions in dimensions and quality occur depending on the device/app to which the book is being delivered. I have not been able to find a way to obtain the best possible quality in all cases.

[JSWolf]

Originally Posted by jhowell

The original images are left intact whenever possible. One situation where that is not the case is when KFX format contains a grid of images that are supposed to make up a single image in the book. In that case the plugin will combine them and deliver the result as a lossless PNG file so that no further degradation of quality occurs.

Be aware that Amazon does not usually deliver the original images as provided by the publisher. Various reductions in dimensions and quality occur depending on the device/app to which the book is being delivered. I have not been able to find a way to obtain the best possible quality in all cases.

I've found a way to get the original images that works when possible. If the eBook is not an exclusive on Amazon, buy the ePub and you'll get the correct images.

[jhowell]

Version 2.16.0 - 06 Oct 2024

Set the PDF viewer preferences direction to R2L if appropriate when converting fixed layout books to PDF format.

Add plugin configuration option "Split landscape images when converting comics to CBZ or PDF" that when checked will cause landscape orientation images in comics to be split into separate left and right side images when converting to CBZ or PDF format. This is intended to break page spreads into individual page images. This option only applies to conversion done using the plugin CLI or From KFX GUI, not conversion using calibre's Convert Books feature.

Prevent failure caused by damaged links between pages when merging PDF files that make up a print replica book in KFX format.

Improve handling of children's books that use PDF images.

Retain filenames from KFX format during conversion to EPUB for HTML files.

[tomsem]

Originally Posted by jhowell

Version 2.16.0 - 06 Oct 2024

Set the PDF viewer preferences direction to R2L if appropriate when converting fixed layout books to PDF format.

Add plugin configuration option "Split landscape images when converting comics to CBZ or PDF" that when checked will cause landscape orientation images in comics to be split into separate left and right side images when converting to CBZ or PDF format. This is intended to break page spreads into individual page images. This option only applies to conversion done using the plugin CLI or From KFX GUI, not conversion using calibre's Convert Books feature.

Prevent failure caused by damaged links between pages when merging PDF files that make up a print replica book in KFX format.

Improve handling of children's books that use PDF images.

Retain filenames from KFX format during conversion to EPUB for HTML files.

Thank you, I will check those first 2 things when I get a chance.

[tomsem]

Originally Posted by jhowell

Version 2.16.0 - 06 Oct 2024

Set the PDF viewer preferences direction to R2L if appropriate when converting fixed layout books to PDF format.

Add plugin configuration option "Split landscape images when converting comics to CBZ or PDF" that when checked will cause landscape orientation images in comics to be split into separate left and right side images when converting to CBZ or PDF format. This is intended to break page spreads into individual page images. This option only applies to conversion done using the plugin CLI or From KFX GUI, not conversion using calibre's Convert Books feature.

These changes will save me a lot of steps for converting to PDF. Thanks!

The only thing I need to check now is that the now-split spreads are paired up properly when the PDF viewer 2 page view option is 'show cover page' (as one would assume is appropriate for any book in PDF format)

Some of the comics KFX's I've authored with Kindle Create have an even number of Single Pages before Facing pages, and when converted to PDF, they need a blank page inserted after 'cover' so the 'spread' pages are paired with each other. They display fine on Scribe (in landscape orientation) because the pages, including spreads, have Single Page property and so display by themselves, while facing pages are always paired up.

But PDF has no concept of 'spread' so I need to insert a blank page to get the pages laid out properly in this case.

At this point I don't know if any comic KFX directly from Amazon ever has this issue, so I won't make a change request for it.


---

Missing from release notes is that From KFX is now setting 'center' property for spread pages when converting to ePub. Thanks for that as well.