too many kerning combinations

Post Reply
Identity-Letters
Posts: 5
Joined: 23 Apr 2021

too many kerning combinations

Post by Identity-Letters »

I tested KO for the very first time today in the morning. The first impression was very good, I am looking for the final version :)

What I found a bit confusing is the amount of Kerning Pairs it generates. If I had kerned the typeface, I expected to get about 1500 pairs. With KO I got about 5500 pairs, but I noticed a lot of combinations I would not have kerned.

Lowercase to Figures
Figures to Lowercase
Uppercase to Figures
Figures to Uppercase
Lowercase to Uppercase (saw that already in another thread)
Currency Symbols to Lowercase
Uppercase to Currency Symbols
(I guess it even kerns more combinations that are not necessary, at least in my eyes.)

Is it possible to have some checkboxes, what KO will kern? Or can I limit KO? Or something similar? The font didn't contain tabular figures, but right now I assume, that KO will kern these as it simply kerns everything?

What I really liked, that KO kerns the exceptions like it generates different values for Vo and Vö. I hated to open the locks… And right now I haven't found any "real" kerning mistake.
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

Thanks for your feedback!

Can you explain why you do not want the combinations you mentioned? Are they combinations that look wrong if kerned?
User avatar
SCarewe
Posts: 25
Joined: 23 Apr 2021

Re: too many kerning combinations

Post by SCarewe »

If I may butt in, just to explain a little, things like uppercase to lowercase are actually essential. Depending on how you space your UC, of course, but I would assume that you've spaced your UC more loosely than your LC, which is very common. Then, though, you'll notice that Hn is a lot looser than dn – as H| is wider than d|. Hence, if you wish to have even spacing between UC and LC, it is necessary to add negative kerning between all of UC and all of LC to some extent. Other solutions include spacing UC like the LC (|H| = |n) and then adding positive kerning between all of UC. This applies just as much to the other groups you mentioned, in order to achieve consistent spacing, at least technically speaking.

On the other hand, I very much see your point and I agree that I would never personally consider kerning things like .n, because it's theoretically looser than n. – something KO will vehemently insist on kerning. I would also very much appreciate a checkbox system as you describe, or at least something that works better than the "Ind." button.

The problem with the Ind. button is that I need to set this for all possible combinations of the same shape with other shapes – for example, if I want to set .n = 0 as an exception, I need to set this for all of lowercase, manually. This is what class kerning is actually for and it would be fantastic to have a little more control. Setting "Independet Models" would be great, so models for specific combinations which don't get taken into account for the kerning of the rest. Does that make sense?
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

I’d recommend to consider the straight sides as “neutral”, e.g. |H| |n u| d|. This is assuming a conventional sans serif design – in serif typefaces the concept of “neutral” sides is much trickier and less relevant.

This means that H. d. and u. should be zero as the period is next to such a “neutral” flat side. Then, having positive kerning for n. makes perfect sense as n| is not a “neutral” side. That way you will also get zero for autokerned .n as you wanted to.
Identity-Letters
Posts: 5
Joined: 23 Apr 2021

Re: too many kerning combinations

Post by Identity-Letters »

They don't look wrong. These combinations are just something I have never kerned. I don't know if they are necessary, right now I can't see a use-case.

I made a comparison of what I expected to receive (1500 pairs) and what I got (5500 pairs).

I am familiar with KernMaster from Dr. Frank Blokland, where I had to create a .txt file, in which I defined, what I want to kern. This .txt took some hours to create, but I was able to use it all the time. it looked like that:

Aa
Ab
Ac


The file had about 12.000 rows, but this gave me control of what I am actually kerning. I think this feeling might be something that also some other users might have. It's difficult to trust a new plugin, when you don't really know what will be kerned. With some checkboxes, this might become obsolete, as you get the impression of actually having more control of the plugin/the outcome.

But anyway, the results are good on the first view, that is what counts.
Identity-Letters
Posts: 5
Joined: 23 Apr 2021

Re: too many kerning combinations

Post by Identity-Letters »

As much as I love, that I don't have to open the locks for kerning exceptions, I just saw, that some accented letters all have tiny different values.

L L -6
L Lacute -16
L Lcaron -17
L Lcommaaccent -17
L Ldot -6
L Lslash -13

same goes for the Letter E or T. It's a serif font, the diacritics are not exceeding the letter. Maybe, these differences could be reduced, when you start rounding the kerning values in the end. I saw that in another thread. For me, I also like to have rounded values, I even have a script (by Lukas Schneider) for that. It's a personal thing, greetings from Adrian Monk.
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

I just posted a video that explains which pairs are kerned by Kern On. Hope that will make you trust the plug-in.

No need to create that text file yourself, it is already supplied with KO (but you could edit it if you really want). Plus, I’m always open for (rational) suggestions.
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

Identity-Letters wrote: 14 May 2021 L L -6
L Lacute -16
L Lcaron -17
L Lcommaaccent -17
L Ldot -6
L Lslash -13
I assume this is not while KO is running but only after finalization, i.e. with class kerning? The classes generated by KO are still not perfect, I will have another look at that part of the code this week.

If LL has -6 then you should probably fix that first, by setting a zero model for LL. This will also tighten the grip on the other L-pairs, and the resulting class kerning will be more consistent.
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

Identity-Letters wrote: 14 May 2021 Adrian Monk.
Sorry, feature requests by Adrian Monk will not be granted.

Keep in mind that we are typeface designers, not font designers. Are you making fonts for the users, or to be looked at in the font editor? There is no point designing a font that looks nice and neat in the editor – the only thing that counts is the final rendering. Btw, the same goes for the “balanced handles” nonsense, another pet peeve of mine.

While I understand that everyone prefers to look at pretty and neat things, and we have to look at our fonts in the editor for many hours, we have to be rational enough to see that the only thing that matters is the rendering of the final font when it’s used.
Identity-Letters
Posts: 5
Joined: 23 Apr 2021

Re: too many kerning combinations

Post by Identity-Letters »

I just tested the latest version, which is way better than the first beta. Congrats on the plugin, I am looking forward to the final version.

I still see some things I don't understand like the Kerning of .tf characters, but this is easy to solve. I will probably use KernOn for an upcoming Lab Release in my foundry. I still have to dig in a bit, but the results were very good :)

Do you have a recommendation, how many Models one should have? In your videos, you have about 70, is that an amount you recommend?
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

Tabular figures (.tf) should not be kerned, of course. That’s simply achieved by setting them to “No kerning” like this:
Screenshot 2021-06-19 at 10.30.28.png
Screenshot 2021-06-19 at 10.30.28.png (27.16 KiB) Viewed 310 times
I will refine the automatic setting of “No kerning” during the initial start-up so as to detect tabular figures and set them to “No kerning” accordingly. Until that’s in place you need to set this manually (just select all figures – you don’t need to do this for each glyph individually).
User avatar
Tim Ahrens
Site Admin
Posts: 140
Joined: 11 Jul 2019

Re: too many kerning combinations

Post by Tim Ahrens »

I will try to make Kern On determine whether it has enough models and maybe provide feedback. You are right, how is the user supposed to know?

Around 70 model pair sounds realistic. Maybe you will add a handful more as you do thorough proofing. You could say, when you don’t spot any more pairs you’d like to be kerned differently you’re done. That said, I will also implement a feature to output a sample text for proofing, with a variety of shapes and special spacing cases, so as to help you achieve some sort of “complete” proof, so you don’t miss any errors.
Identity-Letters
Posts: 5
Joined: 23 Apr 2021

Re: too many kerning combinations

Post by Identity-Letters »

Maybe, you could also make Kern On check, if there are enough models with different letter categories.

Like 15 Models "Lowercase Lowercase"; 15 Models "Uppercase Lowercase"; 15 Models "Uppercase Uppercase"; 10 Models "Lowercase Punctuation"; 10 Models "Figure Figure", … something like that.

I feel, that my focus is set too much on the kerning of Lowercase and Uppercase Letters, so I might forget to kern the other letters. Right now, Kern On also hasn't suggest me to set a kern model for currency figures, symbols like the &, @, ... or a different version of the figures, like osf. Maybe you could add those as a model suggestion.
Post Reply