Font subsetting problem, DOCX to EPUB

[jindroush]

Hello,

I noticed a problem in DOCX->EPUB (AZW3) conversion, but I'm not sure if it should be fixed. But I still think at least some warning would be nice.

The problem is that sometimes you may unknowingly use the font in WinWord which has only REGULAR version, but Word lets you use it with faux bold/faux italic. If you do this and convert to reader format with "Subset all embedded fonts", the same regular font file is embedded and subsetted multiple times, but then, Kindle (at least my Voyage) does use only ONE of the font files for all text using it. Then some glyps as missing depending on which of the subsetted font files the Kindle uses.

See attached files:
test - test.docx is testing docx file. 4 lines with Font1 are in Times New Roman, 4 following lines are with other font, visibly different to each other.
See winword.jpg for how it looks like on my system.

test-test.epub is conversion result with subsetting. It uses correct font and all glyphs for Font2, but not 'faux' versions. I leave it to Kovid if this is a bug or not, but I'd still think this deserves at least conversion warning. But is there anything like conversion warning right now? I know it may be hidden somewhere in conversion log, but some siren and red dialogue would be slightly more convenient

test1.jpg is how subsetted conversion to azw3 looks on my kindle. Uses regular font, but only some glyphs.

test2.jpg is non-subsetted conversion, I assumed it'd be enough for workaround but it doesn't look like it. I though calibre would use font from system, but it probably choses one of the subsetted font files saved in docx file.

So full workaround in this case is no subsetting in word, no subsetting in conversion and accepting huge files (for example Times New Roman is huge). And still, the conversion would not look the same as in winword, because faux-bold/italic is not converted.

BTW: How does the epub look on your reader? Because both Calibre viewer and Kindle Previewer do display it semi-correctly (no faux, but readable), which is slightly concerning...

[kovidgoyal]

That's a bug on the Kindle, if its using the wrong font file, report it to amazon. calibre does not support conversion of fake bold/italic and is never going to. There is no shortage of fonts with bold and italic faces these days that one needs to use fake bold/italic.

[jindroush]

Yes, you're right that it's mostly Kindle bug. But because it's not visible anywhere in DOCX->calibre->epub->kindle chain, it's kind of hidden from the user until the very moment of 'where are my glyphs?' when reading on actual device.

Hence my suggestion about the possible conversion warning. I still think that conversion quirks/known problems/known misses (for example non-conversion of MathML, non-support of small caps, non-support of fake-bold) should be clearly reported to the user. At least it's convenient for the user and prevents future questions/bugreports.

[kovidgoyal]

Except that implementing a warning for a bug in the Kindle firmware means that I then have to keep track of the status of the bug on the Kindle firmware. That's simply not going to happen.

And there are literally thousands if not millions of things that dont convert, and that set of things depends on the fractal combination of the input format, the output format, the users conversion settings, the final reading software the book is viewed on, etc.

[jindroush]

I meant the warning about fake-bold, not about the kindle bug.
But, nevermind.

[Quoth]

Originally Posted by jindroush

The problem is that sometimes you may unknowingly use the font in WinWord which has only REGULAR version, but Word lets you use it with faux bold/faux italic. If you do this and convert to reader format with "Subset all embedded fonts", the same regular font file is embedded and subsetted multiple times, but then, Kindle (at least my Voyage) does use only ONE of the font files for all text using it. Then some glyps as missing depending on which of the subsetted font files the Kindle uses.

1. Don't use such a font.

2. For novels and most writing, LO Writer has been better than MS Word for years. Better than versions 2007 and older for over a decade.

3. You can see in Calibre if a font is missing bold and/or Italic versions and use Font tools to swap it.

4. Download and install fonts that are free and use them.

Note that the 4 oldest Kindles (excepts K3) only do Mobi which doesn't do embedded fonts.

Check the ebook works with a default font on a device as users can't be made to use a Publisher's embedded font.

Put too many fonts and Amazon KDP will remove them.