Overall 'grade' of kerning

Post Reply
User avatar
Jeremy
Posts: 7
Joined: 04 Jan 2021

Overall 'grade' of kerning

Post by Jeremy »

Is there a way to perhaps set KO's approach to the overall kerning – along a scale to let it be 'Timid' to 'Aggressive' (Light to Tight).

I've started following your vid and added the pairs you did and find that KO does like to tighten generally. It would be good to be able to tell it to 'hang loose' a bit and chill out :)
User avatar
Tim Ahrens
Site Admin
Posts: 164
Joined: 11 Jul 2019

Re: Overall 'grade' of kerning

Post by Tim Ahrens »

Agreed, in the beginning, when there are really just a few models, Kern On sometimes does funny things. It should behave more and more as you want as you add more models. After setting 30–50 models you shouldn’t see terribly wrong kerning any more, and then allow for another bunch of models to iron out the remaining deviations from your taste. Plus some models for defining the special spacing, of course.

Thinking about the code under the hood, it would be possibly to have some slider to control ‘Timid’ to ‘Aggressive’ but I wanted to avoid this kind of interface (that’s just the kind of control I had in mind on the belly of the robot). Also, what would happen if you set the slider to ‘Timid’ but your models are kerned the ‘Aggressive’ way, or vice versa?
adrienvasquez
Posts: 2
Joined: 23 Apr 2021

Re: Overall 'grade' of kerning

Post by adrienvasquez »

Hi Tim, I had a thought that echoes a bit what Jeremy is asking about here. I (usually) use ‘round’ values when kerning ie -5, -10, -15 but not -16 or -17 for example. Could there be a way to limit the steps that KO is allowed to use?
User avatar
Tim Ahrens
Site Admin
Posts: 164
Joined: 11 Jul 2019

Re: Overall 'grade' of kerning

Post by Tim Ahrens »

Hi Adrien, can you explain why you would find that feature useful? What’s the reasoning for restricting the auto-kerned values to multiples of 5?
adrienvasquez
Posts: 2
Joined: 23 Apr 2021

Re: Overall 'grade' of kerning

Post by adrienvasquez »

In most cases I don't find the difference between -20 and -21 or -22 for example to be meaningful (while kerning, not fitting) which is why I prefer to work with increments that aren’t too small. I know that not everyone would agree, and I guess I can equally well work like this on my models and let Kern-On figure out the rest, I was just curious about the possibility.
Simon Mager
Posts: 2
Joined: 26 Apr 2021

Re: Overall 'grade' of kerning

Post by Simon Mager »

Hi Tim, I agree with Adrien. Would be very much interested in being able to set the minimum increment to e.g. +/- 5 (not least to reduce kerning data). Thanks for this great tool by the way!
User avatar
Tim Ahrens
Site Admin
Posts: 164
Joined: 11 Jul 2019

Re: Overall 'grade' of kerning

Post by Tim Ahrens »

Thanks, Simon!

We have to distinguish between a minimum threshold, and restricting the steps in which kerning values are set:
  • Restricting the kerning values to certain steps: sorry, I still do not see the benefit of such a feature. What’s the reasoning for restricting the auto-kerned values to multiples of 5? Someone will have to explain again.
  • Having a threshold: There is an internal, hard-coded threshold, you can’t set it, but it is not important in my opinion. Values below a certain threshold will be omitted anyway, while the (class) kerning is finalized. This depends on the kerning size you set in the little dialog.
    In order to “reduce kerning data”, all you have to do is set a smaller figure there. I know it’s not exactly the way of thinking we are used to but Kern On does not minimize the kerning data for a given set of pairs. Instead, it maximizes the kerning it can fit into a given data size. Applying a threshold is part of the algorithm.
User avatar
Tim Ahrens
Site Admin
Posts: 164
Joined: 11 Jul 2019

Re: Overall 'grade' of kerning

Post by Tim Ahrens »

Or, did you mean that restricting the kerning values to certain stems makes the kerning easier to compress in webfonts? That’s something we do with our webfonts at JAF (the code is here). I definitely have that in the back of my head as a future feature, to optimize the whole kerning to get as much as possible out of compressed kerning in webfonts but it’s a real can of worms and it will take a lot more research.
Simon Mager
Posts: 2
Joined: 26 Apr 2021

Re: Overall 'grade' of kerning

Post by Simon Mager »

Hi Tim, thanks for the info. True, the kerning size is handled in the end when kerning on, makes sense to me. I guess the steps are mostly a question of personal workflow. Raising the minimum value to 5 makes it slightly easier to decide in cases of small kerning interventions if a pair needs kerning at all. But I guess you’re right it’s maybe not 100% necessary.
Post Reply