Kerning styles demonstrated

Nick Shinn
Nick Shinn Posts: 2,152
edited June 2022 in Technique and Theory
Nick Sherman’s brilliant implementation of the variable format to kerning, in his new Franklin, features kerning axes that represent two different paradigms. That the end-user may now vary the style of kerning incrementally with both these is really quite profound.

Although, in order to make the concept work, Nick has exercised his taste in putting a huge amount of kerning data into the font, it nonetheless puts the onus on typographers to make their own aesthetic decisions about letter spacing.
In effect, we are entering a third era of non-manual kerning.
  1. No kerning.
  2. Foundry kerning.
  3. Typographer kerning.
Any idea that there is some perfect theory and method of kerning that may be programmed into a font is now dead. More power to the typographer and their taste.

Another possibility would be to put two kerning styles on the same axis, so that the users may slide between them. But perhaps the “Tyght” axis already does that.

nb. It has previously been possible for typographers to apply both methods to the same font, by turning off the “Metrics” kerning in a tight-but-not-touching font, and applying “Optical” kerning—but that lacked the variable dimension.


  • Ruixi Zhang

    I don’t understand the purpose of being offered continuum choices between two foundry kernings, which are so aesthetically distinct. What is “61.8% ‘normal’ kerning mixed with 38.2% ‘tight-but-not-touching’ kerning” supposed to achieve?

    I understand the extreme ends of foundry kernings: The “normal” kerning preserves some kind of rhythm while the “tight-but-not-touching” kerning evokes a particular style and feel, both of which I may or may not agree with the foundry, aesthetically. I tried setting the word “RAVEN” in HEX Franklin—the in-between kerning did not work for me at all.

  • Adam Jagosz
    Adam Jagosz Posts: 689
    I think that specific words not being laid out nicely at specific axis locations do not invalidate the idea, which is exactly to allow setting text, or even single words, to your liking in an efficient manner. I like the separate axes as they seem to give more precise control.
  • Craig Eliason
    Craig Eliason Posts: 1,410
    Agree and agree. Maybe it'd make sense in this case for some axis remapping to hasten the sliders through those less useful middling values and put the precision on the extremes. 
  • Mike Wenzloff
    I suppose as the demo appears to be acting evenly upon all letters at once, it's tracking, not kerning.

    My old person sensibilities to setting type is using one person's pre-programed (yes, but variable) is...meh. In the present demo I feel that just because it can be done doesn't mean it's automatically a good thing.
  • Nick Shinn
    Nick Shinn Posts: 2,152
    An interesting research project:
    Have different type drawers space and kern the same font.


    I know of four paradigms/methodologies:
    • Tight but not touching
    • Tight with overlapping
    • Normal for text
    • Monospace
    Are there others?
  • Ruixi Zhang

    The word “RAVEN” is chosen on purpose—it has an awkward “RA”, a diagonal-to-diagonal “AV”, a diagonal-to-straight “VE”, and a mostly neutral “EN”, in quick succession. This is ideal for evaluating how a typeface performs (I could have gone with the more extreme “LAVENDER”).

    The in-between kerning does have novelty: Typotheque’s Lava comes to mind. In Size-Specific Spacing of Fonts, Peter Biľak writes “[e]ssentially, we spaced and kerned every font in the family three times, for a small size of about 6 points, a default size of 10 points, and a large size of 48 points […]” In this case, the intention of the variable spacing and in-between kerning is clear—I understand that the interpolated spacing and kerning are related to size.

    Nick, I know of 8 paradigms (for proportional fonts) devised by Igino Marini at iKern.

  • Nick Shinn
    Nick Shinn Posts: 2,152
    I find that (8 paradigms) impossible to understand.
  • Thomas Phinney
    Thomas Phinney Posts: 2,770
    It is complex, but completely decipherable. It makes sense as a 2D space. Each typeface getting an iKern treatment, I imagine Igino has to decide where to position it in that space. This is why iKern is not fully automatic, and never will be. (But could well involve just a couple of sliders or dials.)
  • Ruixi Zhang

    I’d envision Igino’s model as a two-axis variation model (2D design space). One axis is “Text-Display” and the other axis is “Legible-Readable”. Every instance is then represented by an arbitrary point within a 2D square. However, within that square, not all instances are considered. Only the instances in the vicinity of an inscribed circle may have the most practical uses (according to Igino). We get the 8 paradigms by cutting that circle into 8 arcs. This is very similar to a color (hue) wheel.

    An interesting observation: Out of the 8 paradigms, only 3 were supplied with illustrations, and these 3 were precisely the ones mentioned by Nick.

    In any case, throwing out less-useful instances (omitting the square’s corners and the disk’s interior) kind of supports my opinions toward in-between kernings.