Kern On ‘flavour’?

Post Reply
User avatar
claegg
Posts: 11
Joined: 23 Apr 2021

Kern On ‘flavour’?

Post by claegg »

Would it be possible to let us users select ‘strategies’ or ‘flavours’ of kerning? Eg. I would like to achieve a picket-fence look. Would such a thing even be possible for some future version of Kern On?
User avatar
Tim Ahrens
Site Admin
Posts: 264
Joined: 11 Jul 2019

Re: Kern On ‘flavour’?

Post by Tim Ahrens »

Kern On guesses (determines) the strategy through your models! That’s the idea: Designers have different tastes and they express this through the models.

Can you describe in more detail what the “picket-fence look” is?
User avatar
claegg
Posts: 11
Joined: 23 Apr 2021

Re: Kern On ‘flavour’?

Post by claegg »

Let me try: Take the word ‘minimum’ as an example. By ‘picket-fence look’ I mean trying to make the stems fall in as regular pattern as possible – like a picket-fence. As an example look at the three images with FFT overlays in the bottom of this article:
http://www.malgil.com/esl/aldus-fft/
User avatar
Tim Ahrens
Site Admin
Posts: 264
Joined: 11 Jul 2019

Re: Kern On ‘flavour’?

Post by Tim Ahrens »

That’s an interesting approach but Kern On works differently: it only considers the white space between the glyph shapes, it doesn’t know about strokes, and it completely ignores the counters.

Kern On live-updates the kerning as you modify the outlines, which allows you to play around with it. You will see that changing the counter shape or size has no impact on the (auto) kerning.

Sorry, the picket-fence principle is not supported by Kern On but, to be honest, I don’t think it leads to better spacing and kerning. Analyzing existing prints with FFT is not enough: We’d have to create one version that is very consistent in terms of frequency spectrum, then another one that isn’t and then see whether the first one looks better than the second.

Btw, similarly, David Kindersley’s ideas on spacing take into account the whole glyph shape, with the centre of gravity of the black etc, and while I think they are highly interesting I don’t think they serve as a model for a spacing or kerning assistant.
User avatar
Tim Ahrens
Site Admin
Posts: 264
Joined: 11 Jul 2019

Re: Kern On ‘flavour’?

Post by Tim Ahrens »

That said, I am aware that in rare cases, the “weight” of the black is not irrelevant. For example, a high-contrast A may have a smaller LSB than RSB even if the outside of the letter was entirely symmetrical. I have had this on my mind as a possible enhancement of Kern On for a very long time but have not implemented this as it would take additional computing time and perhaps make the system over-flexible. (Although Kern On does not employ any sort of machine learning, I am thinking of the concept of overfitting here.)
User avatar
claegg
Posts: 11
Joined: 23 Apr 2021

Re: Kern On ‘flavour’?

Post by claegg »

I see. Kern On only looks at the inter-glyph space, got it.
Post Reply