"Creation" of fonts ex nihilo (or how Calibre displays Bold having only Regular)

[Ghitulescu]

I see that Calibre displays (in preview but also in the standalone, that module that launches when one clicks upon the epub format in the library) the book as if all variants of the font exist (it makes it bold or italic, although only the regular version exists).

Both ADE on computer as the ones used in Sony T series (I have them all three), do not create fake bold/italics. I do not have other ereaders, if one disregards the very basic Trekstor Pyrus Mini.

The fonts are TTF.

I understand that OTF may have some hints/pointers/info helping the viewer display also other shapes (this is how variable width works).

Is calibre doing a service, or there is some "unlocking" magic commands?

Thanks

[Tex2002ans]

If the font variant (or combination) does not exist, some renderers generate a "fake" font.

• For bold, they may take the regular font and make it thicker.
• For italic, they may "tip" or "slant" the letters slightly right.
• For smallcaps, they may shrink ALL CAPS instead.

As an example, see these articles showing:

Real vs. Fake Smallcaps

https://techstacker.com/fake-small-c...al-small-caps/
https://tex.stackexchange.com/questi...xetex-fontspec

(Real smallcap fonts have slightly thicker letters, so they blend in with regular text. ALL CAPS shrunk down becomes too thin, so they look much lighter than surrounding text.)

Real vs. Fake Italics

https://graphicdesign.stackexchange....ustrator/42413

(In many fonts, real italics are usually more "cursive". The fake ones are just slanted.)

Real vs. Fake Bold

https://css-tricks.com/watch-your-font-weight/

* * *

Side Note: These are also called "faux" or or "artificial" or "synthetic" fonts.

For more information, also read the fantastic "CSS Fonts Module Level 4". They give lots of examples + show comparisons + explain in detail.

For example:

2.2.2. Missing weights

Quite often there are only a few weights available for a particular font family. When a weight is specified for which no face exists, a face with a nearby weight is used. In general, bold weights map to faces with heavier weights and light weights map to faces with lighter weights. [...]

Most user agents model a font as having a particular weight which often corresponds to one of the numbers in the nine-step scale described above. While this is true of most fonts, some fonts might be configurable so as to support a range of weights. In this situation, the user agent uses a face with a weight as close as possible to the weight requested [...].

Although the practice is not well-loved by typographers, bold faces are often synthesized by user agents for families that lack actual bold faces.

And especially read "5.2. Matching font styles", which describes each step in extreme detail.

Long story short:

Some programs look for an exact matching/existing font, and if it doesn't exist, will fallback to Regular (or closest thing). (Like ADE not supporting or displaying font-variant: small-caps.)

Other programs like Word/LibreOffice will generate artificial combinations if the font doesn't exist. (Like small-caps -> shrunken ALL CAPS... or if you choose some odd combination like: bold+italic+smallcaps font, it may take the bold+italic font which does exist, convert to ALL CAPS, then shrink.)

[Ghitulescu]

Thank you, it was extremely well-written and exhaustive.

I did some research before but it was hard to filter out the noise and got no real information.

[Hitch]

Originally Posted by Tex2002ans

If the font variant (or combination) does not exist, some renderers generate a "fake" font.

(snippage for brevity)

Long story short:

Some programs look for an exact matching/existing font, and if it doesn't exist, will fallback to Regular (or closest thing). (Like ADE not supporting or displaying font-variant: small-caps.)

Other programs like Word/LibreOffice will generate artificial combinations if the font doesn't exist. (Like small-caps -> shrunken ALL CAPS... or if you choose some odd combination like: bold+italic+smallcaps font, it may take the bold+italic font which does exist, convert to ALL CAPS, then shrink.)

And let me say, for commercial eBookmakers, those damned faux fonts, created by typographers, are no fun either. I just went through this, with SKEWED faces in a goddamned book. Of course, we didn't realize that the bloody thing had faux skewed faces (for, of course, SANS SERIF faces, insert loud expletive deleteds here...) and the customer refused to lose the "precious" sans serif face he had. I explained that we had no ability faux-italicize.

We ended up (I ENDED UP, that is) spending two hours, finding a nearly-identical sans-serif font that, lo, had an italic face (n.b.: you'd be shocked to know how many don't have one) that we could substitute.

I really wish there was some magic button I could push so that I'd have some way of knowing that an italic or bold face was being fake-created by a typographer/designer.

GRUMBLE.

Hitch

[Quoth]

Some fonts even have a slant face and a true italic, because the italic is not simply a slanty font.

[Hitch]

Originally Posted by Quoth

Some fonts even have a slant face and a true italic, because the italic is not simply a slanty font.

Yeah, but, it's beyond rare. Hell, I don't think I recall one.

Hitch

[Quoth]

Originally Posted by Hitch

Yeah, but, it's beyond rare. Hell, I don't think I recall one.

Hitch

I've not seen it. It was just in an article explaining fonts. Some people curiously don't want a real Italic (which sort of was an Italian idea to make type look like script without being a fancy script) but simply want it slanted. Doing slanted or bolding badly on a simple bitmap font* was how early terminals did it. Some small mono LCD panels on instruments or radios still use this using on the fly rendering via the microcontroller. Makes no sense really on a proper font.

( * Such as 5 x 7 or 8 x 8, invariably fixed size and monospace)

[DNSB]

Originally Posted by Hitch

Yeah, but, it's beyond rare. Hell, I don't think I recall one.

I've run into one font which came with oblique and italic styles for the serif, sans serif and monospace versions. Oblique looked better for the monospaced version.

[hobnail]

Originally Posted by Hitch

Yeah, but, it's beyond rare. Hell, I don't think I recall one.

Hitch

Computer Modern had slanted as well as italic. It was a great font family in the sense that Knuth designed a base font with Metafont and then by using parameters was able to tweak them and make all of the different variations. So they all worked together very well. Sort of similar to Adobe's Multiple Masters except that the Metafont fonts were bitmap fonts. There's something newer these days but I forget its name.

[Quoth]

Originally Posted by DNSB

Oblique looked better for the monospaced version.

Yes, it would. Also looks better for small dot matrix bit-mapped fonts than a true italic as there are usually not enough dots.
Oblique is maybe more correct than slanted.

[Hitch]

Originally Posted by hobnail

Computer Modern had slanted as well as italic. It was a great font family in the sense that Knuth designed a base font with Metafont and then by using parameters was able to tweak them and make all of the different variations. So they all worked together very well. Sort of similar to Adobe's Multiple Masters except that the Metafont fonts were bitmap fonts. There's something newer these days but I forget its name.

The thing with Computer Modern, though was that the oblique versus italic seemed sort of...random, if I recall correctly--right?

Hitch

[hobnail]

Originally Posted by Hitch

The thing with Computer Modern, though was that the oblique versus italic seemed sort of...random, if I recall correctly--right?

Hitch

Lucky you that you can remember that far back.

[Hitch]

Originally Posted by hobnail

Lucky you that you can remember that far back.

Only random stuff. :-) Isn't that why search engines were invented???? LOL...

Hitch

[Tex2002ans]

Originally Posted by Hitch

And let me say, for commercial eBookmakers, those damned faux fonts, created by typographers, are no fun either. [...] and the customer refused to lose the "precious" sans serif face he had. I explained that we had no ability faux-italicize.

Woof.

Originally Posted by Hitch

Yeah, but, it's beyond rare. Hell, I don't think I recall one.

Indeed. Very very rare.

(There's also odd/rare combinations, like Upright Italics... mostly used in Maths.)

* * *

And then in CSS, along with your typical font-style: italic, there's even:

• font-style: oblique;
• font-style: oblique 10deg;
  
  • Can go anywhere from -90 -> 90 deg.

but support for this is... crappy. See:

• Mozilla: "font-style"

Originally Posted by Hitch

I really wish there was some magic button I could push so that I'd have some way of knowing that an italic or bold face was being fake-created by a typographer/designer.

"Variable Fonts" are theoretically going to solve some of this—you'll have about 12 different variables you can tweak, like:

• Weight/Width
• Italics
• Slant
• Optical Size
• Grade
• [...]

morphing a single font into all the different variants.

So instead of having 12+ versions (Bold, Italic, BoldItalic, Light/Medium/Heavy, Condensed/Extended, [...]), you'll only have a single font that can be stretched/changed.

If you wanted to play with some of the sliders in different fonts, see:

• V-fonts.com

For an easier overview article, see:

• Web.dev: "Introduction to variable fonts on the web"

For in-depth technical details, see:

• Mozilla: "Variable fonts guide"
• Microsoft: "OpenType Font Variations Overview"
• Microsoft: "OpenType Design-Variation Axis Tag Registry"

But, in many real-life cases, I think it's going to create lots of disasters.

Like those rotten sites that use CSS letter-spacing, but never take into account people who block/override + aren't using their specific fonts. The text clashes together (or becomes too far apart) and becomes completely unreadable.

(Similar to when you export Fixed Format EPUBs out of InDesign—or don't have InDesign's stupid "cloud fonts"—but it still carries over the disgusting CSS, causing chaos in the fallback fonts.)

- - - - -

Side Note: There are huge advantages though with Variable Fonts. Filesize being a huge one.

Most of the fontfaces have nearly-duplicate information, with only minor differences between, so 90%+ of that can be compressed away:

Instead of Regular+Bold+Italic+BoldItalic taking up 2MB + 2MB + 2MB + 2MB = 8MB, it might be a 2.5MB variable font (which includes "all 4" + thousands of in-between variants).

Side Note #2: Most of this "variable tweaking" is also already done by font designers when designing fonts.

There are so many characters in Unicode now, it's impossible for a single person (or group of people) to design a all the typefaces by hand... so a designer will:

• Design the main font/character.
  
  • Like the Regular + Bold + Italic versions.
• Mathematically generate the different weights/variants.
  
  • Tell the computer: "Make all lines thicker/thinner."
    
  • Or tell the computer: "Here's the beginning (Regular), here's the end (Bold), generate 4 different steps between my designs."
• Manually correct any problems.
  
  • Like clashing/overlapping lines, bad kerning, etc.

A lot of this discussion was covered in the absolutely fantastic talk from last year's TUG 2020 conference:

• TUG 2020: "Steven Matteson — The road to NOTO"

which explains how the Noto family of fonts was created over 15+ years.

Noto is an enormous open-source fontface designed to support most of Unicode:

• Over 90 languages + 77 thousand Unicode characters supported
• 72 different typefaces

Originally Posted by hobnail

Computer Modern had slanted as well as italic. It was a great font family in the sense that Knuth designed a base font [...] There's something newer these days but I forget its name.

If you want the original "Computer Modern" font, just expanded to support many more characters/languages...

There's "Latin Modern" font, which was created in 2002 (last updated 2021):

https://ctan.org/texarchive/fonts/lm
http://www.gust.org.pl/projects/e-foundry/latin-modern

and "New Computer Modern", which was created in 2019 (last updated 2021):

https://www.ctan.org/texarchive/fonts/newcomputermodern

- - - - -

Side Note: There are a few more "Computer Modern"-based fonts over the decades, like "CM-Super", etc., but many of these haven't been updated to support lots of the newer Unicode characters + OpenType features.

There are also many other high-quality fonts out there... many are listed in:

• TUG.org: "The LaTeX Font Catalogue"

Side Note #2: And as discussed way above, there are so many more characters (in Unicode) now compared to back then.

To support all possible combinations is a monumental task.

With many fonts, they also tend to be very English- and/or Western-European-centric, so they break (or don't support) writing with things like:

• complicated/unexpected/weird accent + letter combinations
  
  • Czech Č + Ď
    
    • You may be writing a foreign author's name in an otherwise all English book... whoops, that character doesn't exist in your font.
  • the Irish "dotless ı" + the Turkish "dotted İ"
    
    • You want to capitalize or lowercase that? Have fun, heh, heh.
  • or the mathematical "hat" in letters like:
    
    • î
      
      • i + circumflex. The dot in the 'i' disappears and gets replaced.
    • B-hat
      
      • B + circumflex. No language actually has this letter, but could be used in Maths.
• non-Latin-based alphabets
  
  • like Polytonic Greek + Cyrillic

Billions of people also read/write with other, more complicated scripts, like Arabic + Asian languages.

- - - - -

Complete Side Note: Because of this thread, a few days ago I was catching up on Harfbuzz's creator—Behdad Esfahbod.

(Harfbuzz is the fantastic "text shaper" created over the past 20 years, and is now the basis for many programs/OSes [Chrome/Firefox/Photoshop/LibreOffice + Android/Linux]. I referenced it a handful of times in posts over the years.)

He gave talks covering lots of this font stuff:

• 2016: "Behdad Esfahbod - Ten Years of HarfBuzz"
  
  • General overview. Explains how characters get put on the screen, how previous methods failed in "edge cases", and how complicated letters like Arabic/Indic get strung together in computers.
• 2015: "Unicode, OpenType, and Fonts"
  
  • Describes how Unicode + OpenType specs do not match, so there's all this complicated stuff and hacks that nobody knew how to map correctly... so you have bugs in:
    
    • Unicode -> OpenType
      
    • OpenType -> Fonts
      
    • Fonts -> Unicode
      
    mixed with font designers not knowing renderings were wrong (or hacking around the hacks, so the font files themselves were broken)... mixed with each OS interpreting/rendering slightly differently...

Sadly, it seems like Esfahbod's leaving the font business. (Due to how corrupt Microsoft + Adobe have been, and how a few people there have been jamming up and holding back fonts+OpenType for many years.)

He summarized a lot of his 20 years of font research here, leaving the groundwork for future formats + enhancements:

• 2021: "Better-Engineered Font Formats"

And he covered a lot of the font history (+ corruption) in his video here:

• 2020: "Font Wars 2.0: The Untold Story"

Interesting things I learned was the introduction of "Color Fonts". There were 4 different implementations all merged into OpenType at about the same time:

• Apple
  
  • Bitmap image-based.
• Adobe+Mozilla
  
  • SVG vector-based.
• Microsoft
  
  • Flat colors only. Introduced in Windows 8 (or 10?).
• Google
  
  • PNG-based approach.

and no OS supports all 4. Each OS only supports 1 or 2 of these.

This color incompatibility was a big disaster, so now some higher ups in OpenType are very hesitant to add further functionality, shooting down lots of potential enhancements/solutions. (The color solution Esfahbod has been trying to get added is "Color Gradients", which would've compromisingly tackled most use-cases... and again unify OpenType across all OSes just like he did with HarfBuzz.)

[Quoth]

And eReader makers might ignore things anyway. Compare Amazon Kindle 2007 Language support & font support with 1997 Windows, Mac and Linux or even 1987 DOS support (you could do R to L or Hebrew, Cyrillic and Arabic in 1987 on DOS. But Classic Wordstar didn't even do 8 bit ASCII for accents, only 7 bit! Like a Daisywheel.).

Personally I hate the way many websites do fonts. Sheer arrogance and also an issue they don't bother hosting the fonts (possible with Google fonts) but load in real time from Google.