Work methods for non-auto kerning?

helle's picture


I find a lot of you professionals saying that autokern is the work of satan and should therefore not be used under any circumstances. So, how do you people do it? The amount of work a font needs in kerning seems like a LOT... is there an aeasier way than working with every single pair (at least about 60X60=3600 character pairs). Some ideas, please. I´m using Fontlab 3.0

sevenfingers's picture

First of all, a typeface that has good _spacing_ doesn't need as many _kerning_ pairs...
Anyway, there's no other way than the manual hell to get a nicely kerned typeface. not that I know of at least.

fonthausen's picture

For example, you could have a look at the Adobe website, where they have put some (technical) information about typedesign.

Here is a thread abaout the standard adobe kerning:


fonthausen's picture

>>First of all, a typeface that has good _spacing_ doesn't need as many _kerning_ pairs...

that is a dangerous presumption.... It depends on if your type is a sansserif or not, and then on how nice you want to be to international users

sevenfingers's picture

Okey, Monday mornings shouldn't be spent trying to make intelligent statements.

I stand by that good spacing makes the kerning process much easier, bad spacing makes kerning a hell and it does increase the number of kerning pairs you will have to make.

(good spacing) wasn't ment to come off as the solution for getting rid of the chore of creating the kerning pairs...

As for the lack of kerning pairs for international characters... No need to tell me, I'm from Sweden. :)

hrant's picture

Richard is right that good spacing reduces kerning. But it's also very true that many designers (especially "anglos") don't pay kerning its due respect.

It's possible to write a book on your question! And such a book might contain as many questions as answers... But:

If you're starting from scratch, make two lists, one that contains right-hand "exceptions" and one that contains left-hand. An exception is a character (or actually a glyph) whose profile (left or right) is irregular beyond a certain threshold. For example, the right side of the lc "t" is almost always an exception; while the left side of the lc "a" is *sometimes* an exception (think of designs like those of Gudrun Zapf von Hesse). The threshold depends a lot on the level of quality you want, in terms of the number of pairs you'll end up with. After you have the two lists, match up each right-exception member with all the left-exception members, and for each pair decide how much (if any) adjustment is needed.

To me this is the ideal method if you make few fonts. But if you make a living in type design, it's more cost-effective to make a list of pairs (or get one from an expert - just make sure he/she is *really* good) and just check those. But if you do that you always have to stay on your toes - lest you miss something parrticular to the design at hand.

Also note that pairs aren't enough by themselves. Since we generally read word shapes (not individual letters) it's important to test words, of course. But that's a different book.


helle's picture

Thanks a lot, all, and especially you, Hrant. There´s something I can work with...

How about that spacing then? Any good online guides you´d know of?

hrant's picture

You mean the spacing of words?
No guide, but I list the 100 most frequent English words on my own site. Go to the far right of:

BTW, Andy Crewdson once put a great on-line version of Walter Tracy's spacing system. Did that go down with the rest of L&S?


Ramiro Espinoza's picture


In this site you can find *complete* issues about spacing and kerning oriented to digital development of typography.

A great site in my opinion...


Ramiro Espinoza

hrant's picture

Yeah, that's a great page - I'd forgotten about that.

The only thing that bugs me about it is its heavy reliance on control strings. They're certainly a great basic tool (I use them a lot, just like everybody else), but they do contain a danger: real language is not "ooonoonononnonnn", and sometimes control strings mislead. The best example I can think of is the letter "oh": if you look at a well-spaced font you'll notice that its sidebearings are much tighter than a control-string method would generate, and the designer has compensated for the method's shortfall "manually", probably by just going in and tightening all the round characters after having done the control-string method. (The reason rounds need to be tighter is simply because round-round adjacency is rare in actual language). Another example is the most common English word, "the": when you rely exclusively on control strings it doesn't come out very nice.


Ramiro Espinoza's picture

Thanks Hrant!
Please, for a newbie like me,
Instead of the traditional "string spacing method", can you explain us a more professional system to space the side bearings?

Do you begin with string and then you improve the spacing with a list of problematic words? (Like "the" and "oh").

The system of Walter Tracy is only about "strings"?


Ramiro Espinoza.

hrant's picture

Hold everything. What I'm talking about is not "professional" at all - maybe it could be called "academic". In fact it's a waste of time if you're trying to make as much money as possible. But I'd like to think that there are better reasons to design type, or at least to attain a level of quality irrespective of finances.

From what I can tell most people use variations/extensions of Tracy's system, which is indeed based on control strings. But even as Tracy states, the eye is the "final arbiter", which means it's not "illegal" to make the "o" tighter for example. But the eye cannot know linguistics - the *mind* has to help advance the quality of the spacing beyond merely control strings + guesswork.

The way I do my own spacing is actually a diluted version of my "ideal" method, but I think it's still more demanding than what the average designer does.

I do the base spacing with control strings pretty much in the Tracy way, but then I tighten the round sidebearings, and apply certain other adjustements (like loosen the lc "el" - to give it more presence in a word - set the word "million" to see), then I test with the most frequent digraphs and trigraphs, then I test with the most frequent English words (I should do more than English, but...), then I test with a list of rare but "problematic" words. Linguistics also applies in kerning, simply because some pairs are more important than others. That's the gist of it - and I'm really not too happy with it - because I have dreams of what spacing *could* be...


Ramiro Espinoza's picture

Well, I really prefer the "academic" approach to the "take the money and run" approach in type design, ha-ha!

Your advices worth a lot for me. I hope my last font has better spacing and kerning.

I believe a perfect spacing only is possible when you work focused hard in an specific language (in a very specific medium -for example, a newspaper's font-.)

In my view if you work "worldwide" is impossible to fulfil the needs of every language.

graficartist's picture

>ville salervo (Helle)
Thank you for asking this question. This thread has answered a lot of questions that I didn't know how to ask.

>et al.
Thanks you for your responses.


johnbutler's picture

OK, I'm late but you might find this useful:

Anyone who thinks spacing and kerning is an unnecessarily tedious process is right. Computers are designed to automate repetitive tasks. To that end, please read pages 424-436 of the FontLab 4.5 manual and spend a minute calculating the time you will save.

Yes, FontStudio had something like this, but FL's version is truly astounding. Combine that with anchors and glyph permutation and you have an amazing time-saving tool.

I wrote a class kerning script for Robofog once. I sold one copy, then this feature came out in FL. Oh well.

Christian Robertson's picture

This thread has changed my life. I don't know why I have
taken so long to really get in and figure out how to space
and kearn my faces, but on reading this thread this
morning I was motivated to get in and put some of my
faces into fontographer. I'm not there yet, but I'm getting


John: I hope you are right. I have been waiting for such a
tool since I started drawing letters on the computer.

Ramiro Espinoza's picture

In my modest opinion, FL's authomatic kerning features are not so reliable.


Jared Benson's picture

Christian, It's really great to see you around again. Welcome back! Looking forward to seeing more of your work.

Syndicate content Syndicate content