Kerning classes and font generation

George Thomas

Fred Wilson said:

However, I didn't manage to find out how this would affect the kerning?

You wrote that your kern table is 95% diacritics and alternates so I assumed you are kerning diacritics to position them. If you position the diacritics correctly using anchors, no kerning to the affected glyphs should be necessary.

Fred Wilson

I think there is a misunderstanding here (or maybe it's me who doesn't understand). That part of the sentence was actually for the flat kerning table (the one I manually generated) not the class based kerning. Does that make more sense?

George Thomas

No. I too may be misunderstanding, but you mentioned diacritics and kerning in the same description along with the number 40,000 which leads me to think you are applying kerning to some or all diacritics. If you use anchors to position the diacritics, no kerning of the diacritic to its appropriate character should be necessary. There will still sometimes be character-pair kerning, of course, but nothing on the order of 40K.

To help clear things up, if you wish you can email me the .vfb file. My email is in my profile.

[Deleted User]

The user and all related content has been deleted.

Fred Wilson

George, I'm sorry, this thread really flew away far from my original question Which is why we didn't understand each other. It's my fault since I only gave information from time to time and didn't think about giving all the information in the first place.

The font I'm working on currently has 880 glyphs, among which a lot of diacritics and around 350 alternates. I have created 34 kerning classes so far, and 291 kerning pairs. Somehow, I was lead to believe that Fontlab wasn't generating all the kerning pairs or keeping all the classes when I was generating the font file (which is why I opened the thread in the first place). I was very quickly explained that I didn't have to worry about it, and found the appropriate tools to check if everything was ok (and it was). I also suggested I would create a flat kerning table to avoid any potential class kerning issues, and I was also told not to worry about it either. Just to try, I created this flat kerning table, and this is where the 40k pairs come from. Even if it seems like a huge number, I feel like it's not "that" much, considering I have applied what I think I can consider appropriate spacing before doing any kerning, and also considering the fact that several kerning classes have between 30 and 50+ glyphs (diacritics and alternates).

To sum things up, I will keep my class based kerning, and will also stop worrying too much 

James, thanks for the confirmation. I had that feeling after doing some more reading, but I will keep this in mind in case I need it one day!

John Hudson

I too may be misunderstanding, but you mentioned diacritics and kerning in the same description along with the number 40,000 which leads me to think you are applying kerning to some or all diacritics. If you use anchors to position the diacritics, no kerning of the diacritic to its appropriate character should be necessary.

The term diacritic may be confusing. It may refer to either a mark of sign applied to a letter, or to a letter that is carrying such a mark or sign. It helps, in discussions such as these, to distinguish between diacritic marks and diacritic letters. I presumed Fred was referring to kerning the latter (because kerning diacritic marks is insane).

Kent Lew

George — When Fred says “diacritics,” I think he means various accented characters, not the individual diacritical marks themselves.

So the large number of flattened pairs are the result of enumeration of kern pairs between classes that comprise base + accented chars (+ possibly several stylistic variants). All of those accented pairs are what I think Fred meant by “mostly diacritics.”

Fred Wilson

Absolutely, I meant accented characters, I actually thought it was the same thing, but I was obviously wrong.

George Thomas

I think he means various accented characters, not the individual diacritical marks themselves.

Kent — I think you are right. I tend to think in fairly strict terms: diacritics == accent or other modifier letters, standing alone. Glyphs with added components such as diacritics == compound glyphs, that sort of thing. Therein lies the misconception on my part in this conversation.

Thanks for helping us to understand and clear this up.   

Ray Larabie

Ray, so you mean I could keep some classes and expand the rest so there are still some kerning classes but the kerning is still expanded (with less pairs then)?

No but I think you're already on the right track. Do your classes and regular kerning. In Options, set how many pairs you want to expand for flat kerning: they'll automatically generate when you export. I decided on 2500 pairs of flat kerning by experimentation. 2500 covers the basics for my fonts and doesn't make the exported font's file size much larger.

The idea of the expanded kerning is to serve a fallback. It's there so your fonts don't look completely broken in older applications... it doesn't have to be perfect.

Fred Wilson

Oh I see, that sounds pretty nice! I was confused about setting a specific number of pairs, but now I understand thanks to your explanation

Thomas Phinney

George Thomas said:

Fred, if you used anchors for the diacritics I cannot think of any situation where you would need to kern any of them.

George, I think you are talking about kerning the bare diacritics; it seems where Fred writes “diacritics” he is using that as a shorthand for “letters with diacritics,” which as you know do indeed need kerning. Although in some cases that kerning might be different than the kerning for the base letter, sans diacritic.

George Thomas

Thomas, that is correct; Kent pointed that out earlier. I have never before seen anyone refer to a compound accented glyph as a diacritic. To me a diacritic is a mark, accent or other letter modifier; that is why I was questioning his kerning and the number of them. It seems to have been all worked out now.   

Thomas Phinney

Sorry, George, I think I missed jumping to the second page of discussion! So my comment was very out of place. Apologies.