kerning kraziness

[ericwilmoth]

Ugh. Ok, i dig the power this program grants a designer, but having never made a font before i had no clue this kerning thing was going to be an issue.

First i did the auto kerning thing--using the extended script (not that i really understood what i was doing)--and the few pairs it brought up i figured out how to neaten up. Then i noticed in the preview window how uneven things were and manually changed all those...but then i realized just how many other pairs there could be…and as i started looking through the forum i began to cringe at some of the kerning related posts.

On top of this, even after setting MS word to accept kerning it is not working. Then i tried using my font in Illustrator and found out it's not working there, either--though in all honesty i'm not sure if it's supposed to or not...though i would have figured it should.

All i wanted to do was make a personal font for doing my comics...now i'm concerned that to deal with this kerning issue it'll be more hassle then it's worth. Can anyone tell me if kerning can work in illustrator...and if so, how to set it so it does...and if all that pans out, is there a straight forward tutorial for a novice like me to quickly tackle these bazillions of kerning pairs? Otherwise i'll just chalk up the last couple days to a learning experience, buy some fonts and forget about making my own.

[Bhikkhu Pesala]

Kerning is the last thing to do with handwriting fonts. The first thing is the get the side-bearings right.

See this thrread

As you're discovering, font editing is not a trivial thing. Stay with it. You can get good results within a reasonable time-frame, but not in just a few days. And you first have to learn how to use the program. The next version of FontCreator will make manual kerning easier. Autokerning can be adjusted to give more/fewer kerning pairs and tighter/looser kerning by adjusting the white space value. Import from file is the easiest method.

How Long Does it Take to Make a Font?

[ericwilmoth]

Kerning is the last thing to do with handwriting fonts. The first thing is the get the side-bearings right.

I assume by side bearings you mean metrics. If so, I’ve used the auto metrics to take care of that already.

I’ve seen both of these threads before—as well as others like them--and while there’s some useful info there, none of them really answer my questions, especially the concerns about kerning in other programs, like illustrator:?:

As for the act of kerning itself, right now I have 47 kerning pairs that I did on my own (actually, the first few were done the first time I did the automatic kern). Now I’m going back to the auto-kern to get a better feel for how that works and to see if it can speed things up for me...however, that’s not making sense to me.

First, I’m importing kerning pairs from the extended file, and telling it to add new kerning pairs to my existing pairs—which again, is currently 47. Next it says there’s 1156 pairs created from that list. I realize that by playing with the values in the additional options window it will give me more or less pairs in the final stage…but two things aren’t making sense—

--with my values set (200 for white space, 100 minimum, replace if more than 25% and excluding empty glyphs) it produces 95 new kerning pairs…for a total of 117—a difference of 22. That doesn’t make any sense, since I had 47 to start out with.

Fine, fine…

…so I tell it to finish anyway, and when it does, I notice in the preview window that most—if not all—of the kerning I did by hand to get up to that 47 is now undone! WTF

Pardon my acronym, but this doesn’t make sense to me. Somehow twenty five of my pairs are being eliminated.

Naturally, i told it to undo.

What also baffles me is that there are clearly loads and loads of potential kerning pairs that need attention…so why do these lists only have a handful of potential pairs listed? I mean, I know I’m a novice at this whole font creating thing, but I’m shocked—yes, shocked —that there isn’t a more complete listing. I read in the forums (as well as in a book on comic book lettering by DC, where they’re using Fontlab) how people are doing most of this manually…and I’m thinking that is about the craziest friggin thing I’ve ever heard, especially considering the level of technology we are at today.

Can someone please—in addition to all the other concerns I’m seeking answers to in this post—explain to me why there is such limits to the auto-kerning that so many tedious, wrist-wrenching man-hours still need to be spent eyeballing individual kerning pairs? If I were to suggest a feature I’d like to see in future versions, it would be software that could handle testing all potential kerning pairs in one swoop, eliminating the need for such agonizingly monotonous work.

So, again, a summary list of my concerns:
Does kerining work in illustrator and how do I set it to work?
Trying to figure out why setting it to work in MS word on Windows XP is having still no effect
Struggling to comprehend why auto-kerning is eliminating pairs I’ve already painstakingly created
Baffled as to why auto-kerning is not more thorough…and hoping there is a larger list I can use than the existing Kern_extended.txt file

Holding out hope...
Thanks!

[Bhikkhu Pesala]

Does kerning work in illustrator and how do I set it to work?

You tell me. I don't have Adobe Illustrator. As it says in the other thread, not all applications do support kerning. So, if you're relying on kerning to make your font look good, it will only look good in applications that do support kerning if users enable it. As far as possible, design your font not to need it. Lowercase at least should work without kerning. Take a look at Comic Sans or Embassy Script BT, for example, and see how many kerning pairs these fonts have. Do these fonts look OK?

Trying to figure out why setting it to work in MS word on Windows XP is having still no effect

The setting also depends on the font size. Usually word-processors will enable kerning only above a certain point size. Does kerning work as expected in other fonts like Times New Roman or Palatino?

Struggling to comprehend why auto-kerning is eliminating pairs I’ve already painstakingly created

First you should do auto-kerning, experimenting with different white space values until the results are mostly OK. Then you have to manually adjust those pairs that are not right.

Baffled as to why auto-kerning is not more thorough…and hoping there is a larger list I can use than the existing Kern_extended.txt file

Kern extended includes lots of accented characters and Greek. Does your font include these glyphs? You may find that the kern_standard.txt file is sufficient.

Designing an algorithm that works well for all fonts is not a simple programming task. It is also a matter of personal taste which pairs should be kerned and by how much.

You can export a list of pairs from one font as a text file and use it for other fonts.

First, I’m importing kerning pairs from the extended file, and telling it to add new kerning pairs to my existing pairs—which again, is currently 47. Next it says there’s 1156 pairs created from that list.

--with my values set (200 for white space, 100 minimum, replace if more than 25% and excluding empty glyphs) it produces 95 new kerning pairs…for a total of 117—a difference of 22. That doesn’t make any sense, since I had 47 to start out with.

There are 1156 pairs in the file, but most of those glyphs do not exist in your font, so only 95 new pairs are produced with a total of 117. Since you already had 47, then only 70 of those 117 are completely new pairs, the remaining 25 are new values for the existing pairs, and 22 pairs will retain their old values because the difference is less than 25%. To avoid replacing old pairs, increase that 25% value to 1000%.

[Bhikkhu Pesala]

Without doubt there is room for improvement in the auto-kerning feature, but you're rather missing the point. You're trying to run before you can walk. First design the glyphs, then manually adjust the side-bearings using the Preview Toolbar and/or glyph comparison toolbar. Autometrics is just a quick and dirty method. There is no way that a font needs the same side-bearings on A or W as it does on i, j, or y. Look at some existing fonts that are similar to yours, and see how they compare.

[ericwilmoth]

There are 1156 pairs in the file, but most of those glyphs do not exist in your font, so only 95 new pairs are produced with a total of 117. Since you already had 47, then only 70 of those 117 are completely new pairs, the remaining 25 are new values for the existing pairs, and 22 pairs will retain their old values because the difference is less than 25%. To avoid replacing old pairs, increase that 25% value to 1000%.

Thanks, that explanation makes sense. I did that and it worked…though after running a few more tests I think somehow the first time I did it I must not have selected the option to include existing pairs like I thought, because trying it both ways I discovered that it only overwrites my manually set kerning if I don’t include the existing kerning pairs.

As it says in the other thread, not all applications do support kerning. Lowercase at least should work without kerning.

Well, I’ve designed my font so that the lowercase are identical to the uppercase, except for the ‘I’ (it’s a comic book thing)…and after some more testing I discovered some interesting things. First, Illustrator does support kerning…but I was getting confused because pairs like ‘TO’ were showing up with large gaps despite me having tightened them up already. What I missed was that the two kerning pairs I’d created were both using the capital ‘T’ –meaning I’d still need to go back and make two more pairs to include the lower case ‘t’ (which in this font appears identical).

Is there by any chance a way to automatically duplicate the kerning from the caps to the lower-case characters

The setting also depends on the font size. Usually word-processors will enable kerning only above a certain point size.

I noticed the concern over font size and that the setting can be modified, and it still didn’t appear to work at any size when I tested it on Friday…but I think that was because of two factors—first, because of the problem I mentioned above with pairs like ‘TO’, and because last time I just typed random sentences, not the whole alphabet from ‘A-Z’…so I didn’t really see my manual kerning in action. Today I did both, which was part of how I discovered the ‘TO’ blind-spot I was encountering.

First design the glyphs, then manually adjust the side-bearings using the Preview Toolbar and/or glyph comparison toolbar. Autometrics is just a quick and dirty method. There is no way that a font needs the same side-bearings on A or W as it does on i, j, or y. Look at some existing fonts that are similar to yours, and see how they compare...and see how many kerning pairs these fonts have. You can export a list of pairs from one font as a text file and use it for other fonts.

I opened a number of fonts, and none of them had any kerning pairs, which I found to be interesting. I looked at their metrics and saw that they varied in width from character to character…but each font was different in this regard. I can see now that modifying the metrics before moving on to kerning will reduce the number of potential kerning pairs…but what I don’t have a sense of is the type of methodology or pattern to doing this correctly. You mention that it is known that certain characters need more or less than others, but how is that determined
Is there a rule of thumb to follow, and if so, what is it

I see that by using the comparison window I may preview how each character rests next to the others…but I feel like I’m missing the bigger, intuitive picture of how to wisely use this tool when adjusting the metrics for a given character to minimize the need for kerning, later. At this point about the best I can do is screw around and try things, but I get the sense that there are some basic, tried and true techniques you pro’s are using that are not plainly stated anywhere in the documentation that a newb like me would find immensely useful.

Part of my problem may also stem from using all caps (except for I), introducing a component I’m too naïve yet to fully appreciate. Any ideas on that

So, if you're relying on kerning to make your font look good, it will only look good in applications that do support kerning if users enable it. As far as possible, design your font not to need it.

What I’ve noticed is that if a font is designed so that the outlines of the characters adhere to a regular rectangular shape then when the characters are lined up end-to-end they are going to look mostly ok, and the standard kerning would probably handle most of the needs for such fonts. The bold version I designed isn’t as rectangularly uniform, which is part of my conflict…and I’m thinking at this point—though I’m not certain—that by skewing it I’ve exacerbated the apparent distance between certain character pairs.

Since there does not appear to be an a-z tutorial on font making either in the manual or in the forum and it’s things like this that are not immediately intuitive, if you or anyone else can share some insider tips on how to intelligently and wisely adjust the metrics to make a font look as good as possible prior to kerning I would appreciate it immensely.

Thank you for all your help thus far!

[Bhikkhu Pesala]

Is there by any chance a way to automatically duplicate the kerning from the caps to the lower-case characters

If the lowercase are identical to the uppercase then they can all be compostes of the uppercase.

• Copy capital A in the glyph overview
• Paste it into lowercae a to copy the glyph metrics
• Make it empty
• Open the Glyph Edit Window
• Paste the capital A into lowercase a as a composte glyph member

Now you only need to adjust capital A. Do all the adjusting of spacing for capitals first, then create composites for lowercase. Kerning can come later if you do need it. The lowercase "i" will have to be a simple glyph as its letterform is different to uppercase "I".

You mention that it is known that certain characters need more or less than others, but how is that determined
Experience: Offhand, I think the pairs that need kerning in most fonts are AT, AV, AW, TA, VA, VW, LT, LV, LW, FA, PA, then perhaps AO LO, OT, etc. If a font has different lowercase, then one also needs to kern Ta Te To T, T. V, V, F. F, etc.
Is there a rule of thumb to follow, and if so, what is it
Running the Autokerning Wizard will quickly show you the pairs that need kerning. The Wizard tends to be overaggressive on some pairs, but not on others. C- AV, T- AW AT can be much too tight.

I see that by using the comparison window I may preview how each

I get the sense that there are some basic, tried and true techniques you pro’s are using that are not plainly stated anywhere in the documentation that a newb like me would find immensely useful.

I'm not a professional, just an experienced amateur. I've been using FontCreator for a few years, and I did some font editing in Corel Draw 3.0 and later in Fontographer.

Part of my problem may also stem from using all caps (except for I), introducing a component I’m too naïve yet to fully appreciate. Any ideas on that
Just focus on getting the Caps right. Since the lowercase are the same, you can ignore them until the caps are finished.

What I noticed is that script fonts don't need kerning as much as regular fonts, but they do need care with side-bearings. It is like the difference between handwriting and lead type.

Try putting a rectangular glyph like "N" or "H" with equal side-bearings in the before and after sides of the Comparison Toolbar. Than scroll through the font seeing how the other letters look between a pair of capital "N"s. For figures, do the same with a pair of zeros. Use the Preview Toolbar to see the wholle alphabet at a glance, and see if any glyphs have too much or not enough space. Make sure that you turn kerning off while doing this.

[William]

I have just produced a font for testing whether an application package is using the kerning information of a font at a particular size.

The font is named AV kerning test of packages. It is available as a free download.

http://www.users.globalnet.co.uk/~ngo/AVKERNED.TTF

Once the font is installed, simply start an application package, select the font and key the following.

AV

The two characters are designed so that if the kerning information of the font is being used by the application package, then the display shows the kerning effect very clearly.

The font only has A, V and space as characters. Keying other characters simply displays the .notdef glyph of the font.

I have tried the font with Microsoft WordPad, where the kerning information was not used, and with Serif PagePlus 10 where the kerning information was used at larger sizes. The size above which auto kerning takes place is 16 point by default and whether to use auto kerning and if so above which size auto kerning is to be used can be set on the Format Character... Character Spacing panel.

William Overington

2 July 2007

[ericwilmoth]

Thanks, especially for the tip about creating composites from the caps. It sounds, though, from your instructions like it doesn’t necessarily utilize the kerning data from the original…or do you know for sure

Incidentally, I discovered something odd, maybe you’ve encountered this…

Earlier, I started using MS Word with character strings like…
AAABACADAEAFAGAHAIAJAKALAMANAOAPAQARASATAUAVAWAXAYAZA1A2A3A4A5A6A7A8A9A0A:A;A,A.A!A?A@A#A$A%A^A&A*A(A)A+A-A=A{A}A|A\A/A>A<A~A’A”A
…and so forth, to search for potential kerning pairs.

I finally realized that by using the autokern wizard and specifically selecting characters to generate kerning that the wizard would generate many more kerning pairs—several hundred to a couple thousand, depending on the number of characters I manually added, of course. At first I thought I’d stumbled onto the mother-load solution…but there’s still a problem. For some reason, even after these pairs are generated, when I reinstall the font the kerned pairs do not change in MS word.

To paint the full picture, i’m doing this with a test version of my font, so as not to pollute my good version; the MS word text is already selected to properly reflect kerning pairs, and I know it works because if I select the text and switch to my primary font, the kerning pairs I’ve programmed are still there…and if I make changes to a pair and reinstall that good font, naturally, the changes take effect—but not so with this test font when I use the autokern to add in the hundreds of other kern pairs.

I tested this a step further by clearing the kern pairs in the test font, then creating one new pair—‘AB’ and giving it a huge gap between them. when it loaded, the gap was there. Once more I did the autokern to produce about 400 kerned pairs, reloaded the font, and once again, none of them appear to be working—not even ‘AB’.

Lastly, I cleared all the pairs again, went back to the wizard, but this time only selected ‘A’ and ‘B’. The wizard then produced a single kerning pair of ‘BA’. I reloaded the font and indeed the characters were updated in MS word. One last test, I opened that pair up and manually inserted a huge gap, reinstalled it again, and yes, the gap was there.

So, the problem seems to be that even though the manual says there can be as many as 10920 kerning pairs (pg 57), for some reason, after a certain number (between 120-400) the kerning pairs stop working. This, to me, either means there’s a bug in the program, or maybe if I uninstalled and reinstalled the software it might fix the problem…but I’m curious to know if anyone else has experienced this problem. If it actually worked, then despite the over-aggressiveness you mentioned about the auto-kern wizard earlier, the darn thing should actually do what I was hoping the system should do, which is deal with all this kerning more thoroughly, automatically. If this is a bug, then I want to be sure to report it, but if it should be working then I want to find out why it’s not. I’ll snoop around and see if I can find a more appropriate place to post this concern, but if you have any suggestions that would be great.

Thanks!

[Bhikkhu Pesala]

The problem seems to lie between the keyboard and the chair.

Selecting A and B will create two pairs, AB and BA, but since AB are usually very close no pair is created for that.

I will not reply again for a while. You need to spend more time using the program before filing bug reports. You are just too impatient.

[ericwilmoth]

Um…I’m confused. Did I offend you or something? I thought this was a forum to get help and advice…so if I find something I don’t understand I’m inclined to ask for help. If you don’t want to help that’s your business…but why insult me and my ignorance when I’m simply trying to learn?


I may be seeking to complete my font as quickly as possible, but I don’t believe that equates with impatience. I performed what I feel is a thorough test before posting this last concern…maybe I’m missing something simple and logical…but as far as I can tell I’m stumped. I don’t know if this is a bug or not...
hence asking questions.

I did opt to post a thread in the bug reporting forum, just in case: viewtopic.php?t=1807
If doing this is somehow taboo, then please point me in the right direction for posting etiquette; I thought this would be a natural and healthy thing to do, not impatient. How else does one discover and or deal with bugs?

If there’s anyone else out there who’d be kind enough to lend a hand I would really appreciate it.

[William]

I have not done much with kerning. I have been trying to follow this thread.

Earlier this morning I started what was intended as a test font for my own use for learning about kerning.

I copied the A and the V from my AV kerning test for packages font and the O from the disc of my Stardisc font.

I added A V O T P R I J K B in that order, adding kerning pairs manually as I proceeded as soon as I had added another glyph, making notes recording each stage in a WordPad file as I proceeded, which is my normal practice when producing a font. Realizing that as the number of glyphs increased that kerning would take longer and wondering about using the autokerning facility later, I then completed the alphabet without adding any more kerning pairs so that I would have a font which could be used for titling.

All of the kernable pairs for A V O T P R I J K B are thought to be complete. Other kernable pairs are not started.

In producing the designs for the glyphs, the A and the V and the O lead to my trying to produce a font with an Art Deco look.

Anyway, the font is not complete as it only has capitals A through to Z and space and kerning is only partly done.

However, I have tried the font using Serif PagePlus 10 for the word OAT and the kerning words well.

So, I thought that I would upload a copy to the web in the hope that the font might be of interest to some readers and might stimulate some discussion of kerning in this thread.

http://www.users.globalnet.co.uk/~ngo/KERND001.TTF

The idea of the 001 is that this is the first version and I will not alter that file. If I proceed I will save a copy as Kern Deco 002 in KERND002.TTF and then edit that and so on.

As I say, I have not done much with kerning, so I too am hoping to learn.

William Overington

5 July 2007

[William]

I have now produced a KERND002.TTF font and added a copy to the web.

http://www.users.globalnet.co.uk/~ngo/KERND002.TTF

This version has the glyphs for the capitals copied to the lowercase positions and has some more kerning pairs added for the capitals. There is no kerning for the lowercase at all. This facilitates both unkerned pairs and kerned pairs versions of how the glyphs fit one against another being compared and contrasted.

There are also some of the digits and some of the punctuation.

William Overington

5 July 2007

[William]

The next version, Kern Deco 003 in KERND003.TTF is now available.

http://www.users.globalnet.co.uk/~ngo/KERND003.TTF

There are some more characters, including É and Ç and lowercase mapped copies as well.

The following kerning pairs have been added, where OE refers to the OE ligature.

AOE
KOE
OW
SA SO
TS
VOE
WO WS WOE
YS
ÇA

I have tried the words CAFÉ and PROVENÇAL in Serif PagePlus 10.

William Overington

6 July 2007

[William]

I produced a later version of the font with some more kerning pairs yesterday afternoon. I have added it to the web earlier this morning.

http://www.users.globalnet.co.uk/~ngo/KERND004.TTF

This adds some more kerning pairs.

Here are links to some test documents which I produced.

http://www.users.globalnet.co.uk/~ngo/k ... 03test.PDF

The above document tries out the two accented capitals in the font. Please note the kerning of the ÇA in the word PROVENÇAL.

http://www.users.globalnet.co.uk/~ngo/k ... 3test2.PDF

http://www.users.globalnet.co.uk/~ngo/k ... 4test2.PDF

The above two documents have the same text yet displayed using two successive stages of development. Please note how several unkerned pairs in the first document are kerned pairs in the second document.

William Overington

7 July 2007