dwdjmu's picture

I want to find out how you guys kern. what do you think makes for good kerning? Do you look for specific problematic pairs of letters, or do you try to make the entire words themselves have even "coloring" or weight? Do you treat words first, and then their relationship within sentences?

I need some experienced advice as how to best approach this.


Mark Simonson's picture

Are you talking about kerning existing fonts in graphics apps or fonts you are developing in a font editor?

Assuming it's more likely the latter, here's what a lot of type designers do, briefly:

Once you have your basic spacing set and things are looking good with unproblematic glyph combinations, you go through and find pairs of glyphs with problematic right and/or left sides (A, L, T, etc.) and then put these in your kerning/metrics editor amongst "normal" or "control" glyphs (usually runs of H and O) and add kerning until you get even spacing or a reasonable facsimile. As a reality check, generate and install the font and set some words, sentences, paragraphs, etc. in a graphics program and go back to the kerning editor if necessary.

It's a pretty tedious task, so it's good to be as systematic about it as you can (e.g., make lists or tables of possible pairs, etc.) and take advantage of tools and shortcuts available (e.g., Kerning Assistance in FontLab and FOG, class kerning, etc.). The thing NOT to do is just dive in and start kerning pairs at random or to attempt to do kerning before the basic spacing is worked out. I also don't recommend auto-kerning.

John Hudson's picture

For each script for which I design, I build a set of spacing/kerning test words, ideally one example for every possible combination of letters. Note that this usually means using words from multiple languages: some Latin letter combinations are only likely to occur in transliteration of Arabic or other semitic languages. These words are sorted according to the pair they exemplify (aa, ab, ac, ad...) and I produce one version, using glyph names, for use in the FontLab metrics window, and one version using actual characters for making test documents in InDesign or Word. For bicameral scripts, I will typically have five kerning lists: lowercase-to-lowercase, uppercase-to-uppercase, smallcaps-to-smallcaps, and uppercase-to-lowercase and uppercase-to-smallcaps.

I generally space a type as I'm designing the glyphs, rather than as a separate step between outlines and kerning. I adjust the spacing as I go along, so that by the time I am ready to start kerning I have the spacing largely how it should be. Even so, it isn't unusual for me to adjust the basic spacing of some glyphs once I start kerning.

I import my kerning lists into FontLab's metrics Window, and start working my way down the list. I generally concentrate on the specific pair that the word is included to present, but will also kern other pairs in the word if they are in obvious need of attention. This is helpful, because when I encounter them again further down the list I can confirm whether the kerning is good.

Basic smallcap-to-smallcap kerning can often be derived from the uppercase-to-uppercase kerning values, so I usually export the latter as an AFM, globally adjust the values in a spreadsheet, change the names to reference the smallcaps, and import the values. Often, only about 20% of the imported values need to be adjusted to account for, e.g. difference of angle between uppercase and smallcap diagonal strokes.

After I've finished kerning, I generate a font, and print out my test word document from InDesign and Word. Then I go through it with a red pen marking pairs that seem too tight or too loose and that need adjustment.

Recently, in some Arabic projects, I've engaged in OpenType contextual kerning. This is not for the faint of heart.

as8's picture

I did kerning in the wrong way because I did not fix
the sidebearings properly.
I went to the cinema to watch this movie,
'Paris, Texsas' (1984) :


Miss Tiffany's picture

Question: How does everyone approach kerning?

Syndicate content Syndicate content