Kern On pair and word lists

James PuckettJames Puckett Posts: 1,849
edited June 9 in Technique and Theory
Has anybody shared a list of model pairs and words used as models for Kern On? I tried searching GitHub but I just get a slew of unrelated stuff.


  • John HudsonJohn Hudson Posts: 2,384
    edited June 9
    You don’t really need a shared list. You always start Kern On by defining a minimum of ten pairs that do not need any kerning, so typically HH HO OH OO etc. for Latin. Then Kern On starts suggesting potential model pairs to look at. You can skip (Next) the suggestion or click the arrow to display the pair, which will show with the auto kerning that Kern On applies based on the current model set. If you edit that kerning, the pair automatically becomes a model. You can also type/paste any other pairs or word lists or blocks of text, and see how Kern On is kerning based on the current model set.

    The model set for a Kern On project tends to be design-specific, but once you have about 50 model pairs you are probably in a pretty good situation for Kern On to do a good job. Because the auto kerning is iterative and you can preview the results in any glyph string, I find I don’t always use the same model pairs, but end up tailoring the set to the nature of the design and where I feel Kern On most needs hints about what to do.
  • jeremy tribbyjeremy tribby Posts: 110
    I saw you ask this on twitter first, but to expand a little: one of Kern On's strengths is that it is not super opinionated about this. Kern On's suggestions are based on pair frequency, but those pairs could be duplicative, for example o-n and o-r are common pairs according to Kern On; it's up to you to decide that o-n is the right one to define, and a model without kerning at that. I think it's helpful to think of Kern On suggestions as considerably more useful after you have made your own decisions about key pairs. so in a sense, there isn't a Kern On-centric pair list, just very good suggestions based on how you think about some of your own pairs.

    (for the sake of others here — to see the pair frequency list, right-click the plugin file itself, select "Show Package Contents" and navigate to Contents/Resources/pair_frequencies.txt — I didn't see anything in the license against viewing that file, but please correct me if I am wrong, @TimAhrens )

    In a font with just under 2,000 glyphs across latin and extended cyrillic, I have found that around 100-150 models per master (including user-defined auto kerned pairs and independent pairs) is where Kern On will stop making suggestions, but I would guess fewer is actually better

    Tim runs a forum for Kern On if you find any of the previous questions and answers there helpful, I know I did :smile:

Sign In or Register to comment.