TypeFacet Autokern

2

Comments

  • Michael ClarkMichael Clark Posts: 138
    edited June 2012
    "If you would put enough effort in it, you can easily match or outmatch human kerning."

    Huh?????

    Surely you jest!
  • RalfRalf Posts: 170
    Of course they are. Humans can have original ideas.
    But how much about the creation of type is about “ideas”? You can think of a type design idea as the definition of a (depending on the typeface simple or pretty complicated) set of rules. And then 99 % of the work is “just” a craft—applying these rules to different letter skeletons, widths and so on.
    Kerning is defining a certain rhythm of stems and white space—as an experienced type designer you just “see” the right kerning because you do the math almost automatically. But that's not about “original ideas” and that's nothing you can't teach a machine. What you do visually is just math. A computer can easily do it if you feed it with the right algorithms and set the right parameters.
  • Whenever I used iKern, I would first do my best with drawing/spacing, even do my own kerning classes and kern, then hand it over because, simply put, Igino was better and more thorough than me.
  • A computer can easily do it if you feed it with the right algorithms and set the right parameters.
    If someone can prove that, I’ll be delighted. But so far all auto-stuff I’ve seen including this one we talk about is not very impressive if you look at it at from a professional level.
  • Perhaps the parameters are too complex. A pile of sand being built one grain at a time until the pile reaches critical instability and collapses cannot be modeled with a computer. The relationships between the grains of sand are too complex and change too randomly for the model to be built.
    Kerning can be described as a balancing of a volume of space between letters. Shape, slope, interior space, length of serif, arc of curve, all of these are (or should be) subtly different in different glyphs. A serif one unit longer than another, will affect the overall volume between letters. The angles of the various diagonal forms are all at slightly different pitches, forming relations ships that may be too complex for a computer algorithm to master. But to a type designer's experienced eye, these things all become obvious.

    Or not, maybe I'm crazy.
  • Dave CrosslandDave Crossland Posts: 1,053
    "The thing I don't understand is the idea that spacing would be a completely separate step."

    That is why I think having auto-anything in the hands of the designer is important - it can be fully integrated into their design process.
  • Nick ShinnNick Shinn Posts: 1,537
    I could certainly have used a kerning application to space Softmachine, in which the closest proximity between every glyph combination is equalized.

    So I don’t deny that a kerning tool might be creatively integrated into the design process of certain type design concepts.

    However, in general it seems to me that spacing is a complex, subjective matter of taste, rather than something which can be “parametricized”. Not every glyph shape can be put into a class.

    I also question whether kerning is pro-readability. Sure, the obvious huge gaps need to be fixed, but below a certain threshold, might not sidebearing space be considered an inviolable part of the letter? Consider the figure “1”: it bothers me when it is fitted, and I give the proportional one more and more space these days.

    Especially for serifed type, the fundamental nature of its construction from Jenson on is to avoid kerning. If you fuck with that, you better be sure of your ground.

    While automation is useful, it can suck the life out of art, rendering it crude and banal.

    I like to dive deep into the zone when I kern, and submerge my being in the spatiality of the typeface, in an organic, holistic way, for hours on end. Feel the space. Can you get as deep interelating with parametrics?

    I would imagine setting up a number of different algorithms for kerning a face, and seeing which worked best, then tweaking it. A quite different MO.
  • James PuckettJames Puckett Posts: 1,699
    Sure, the obvious huge gaps need to be fixed, but below a certain threshold, might not sidebearing space be considered an inviolable part of the letter? Consider the figure “1”: it bothers me when it is fitted, and I give the proportional one more and more space these days.
    I don’t recall anybody saying automated kerning has to work like Ed Benguiat circa 1975.
    Not every glyph shape can be put into a class.
    I just did a custom job on a tight deadline and used iKern to get the job done quickly. The really irregular shapes—mostly punctuation—had to be fixed manually. But fixing those oddballs took less than an hour, most of which was spent looking over a proof to find the problem. I can live with that.
    While automation is useful, it can suck the life out of art, rendering it crude and banal.
    Just like the Benton matrix cutter ruined type design and that’s why all those fonts by Goudy and Benton are so awful.
    I like to dive deep into the zone when I kern, and submerge my being in the spatiality of the typeface, in an organic, holistic way, for hours on end. Feel the space.
    I didn’t realize pot was legal in Toronto!
  • Max PhillipsMax Phillips Posts: 464
    I would imagine setting up a number of different algorithms for kerning a face, and seeing which worked best, then tweaking it. A quite different MO.
    Actually, I think that's pretty much the MO the pro-automation posters have been talking about. No one is suggesting there's a single correct way to kern—most people seem to be discussing algorithms with adjustable parameters—and no one's suggesting the designer should accept the algorithm's output without reviewing and fine-tuning by eye.
  • Nick ShinnNick Shinn Posts: 1,537
    edited July 2012
    Just like the Benton matrix cutter ruined type design and that’s why all those fonts by Goudy and Benton are so awful.
    I said “can” suck the life out of, not “always does”.
    Obviously, automation is useful, but on the other hand there are faux italics and faux small caps, which you presumably think are great, because Morris Benton used automation? As for Goudy, he was equivocal: “Punch cutting by hand versus matrix engraving by machine is a subject for dispute.” (Typologia, p.106).

    Automation is not a magic bullet that kills every badness it touches. Sometimes it just kills indiscriminately.
    I didn’t realize pot was legal in Toronto!
    Pot is great for creative ideas and improvisation, but the short term memory-loss it produces is not useful for production work, especially the kind of whole-font spatial awareness employed in kerning.
    …I think that's pretty much the MO the pro-automation posters have been talking about. …
    And that’s what I have been criticizing, as being both a superficial appraisal, and redundant.
    After all, if you’re going to go through every kern pair to check that they’re all correct (e.g. Ray’s mention of Oslash), and tweak those that aren’t, how is that different from class-based kerning with comprehensive inspection and manual tweaking that I do now?


  • Just like the Benton matrix cutter ruined type design and that’s why all those fonts by Goudy and Benton are so awful.
    Goudy didn't use a Benton cutter. Originally all his types — with one exception being Goudy Oldstyle which he sold to ATF and then complained when Benton kept making additions and alterations — were cut by Robert Weibking using engraving machines of his own design. Goudy bought one before Weibking died and after that cut his own matrices. Weibking also cut the Centaur type for B. Rogers using the same machines.

    George
  • Why don't we just automate the whole process? We could all get together and write a program that would design fonts automatically, we can build in all possible parameters and then let it run. If we could turn it into a virus then every computer in the world could turn out more fonts than we would ever need. We could give up type design and apply our skills to solving bigger problems like world peace. Or maybe just write an algorithm for that too.
  • Just wanted to add that Weibking also furnished and trained the engraving department at Ludlow.
  • Just wanted to add that Weibking also furnished and trained the engraving department at Ludlow.
    Thanks for the reminder!
  • Ramiro EspinozaRamiro Espinoza Posts: 764
    edited August 2012
    My 2 cents on the issue of auto kerning: I think the problem (as in many other subjects) is speaking in absolute terms. For example, I've been testing the new Kern Master and I think the results are pretty good. I wouldn't leave the kerning it generates as it is BUT if you check and correct with your kerning pairs lists after the software did its work, you save quite a lot of time. In some cases more than 50% of the kerning can be left as it was done by KM. There is a huge difference between making the kerning from scratch and just correcting it here and there.
  • Ben BlomBen Blom Posts: 248
    I am afraid I do not agree with some other contributors here. I would like to share some thoughts related to auto-kerning and auto-spacing.

    In the context of designing an application for auto-kerning and auto-spacing, I would suggest the following general relationship between glyph design, spacing, and kerning:
    Design all glyphs of font —> create provisional spacing —> create provisional kerning —> create final spacing —> create final kerning.

    Spacing and kerning should follow from glyph design, not the other way round. So, a specific glyph design needs a specific spacing and kerning, to make it look good in the context of other glyphs — whatever that spacing and kerning may be. It's not like this: a designer first decides what spacing and kerning to use for the "R" glyph, and only after that, he designs the "R" glyph. Because there is a relationship between the design of a glyph, and the spacing and kerning which one thinks should go with that glyph, a typeface designer can take this relationship into consideration when designing a glyph — but that doesn't reverse the direction of the relationship between a glyph design, and the spacing and kerning which should follow from that glyph design.

    So, when a designer designs an "R" like the one in Gill Sans, he should know that it will have different spacing and kerning implications than designing an "R" like the one in Trade Gothic. And this designer can, of course, create any provisional spacing (and even kerning) to see how the "R" he is designing, functions in the context of other glyphs. And when the designer then discovers a so-called "spacing problem that is caused by the shape of a letter", he can fix that. And of course, this can be an iterative process of changing the shape of a letter and changing the provisional spacing of that letter (and changing the shape/spacing of other letters). A process like this can lead to the final design of the letter involved, but it will definitely not lead to the final spacing for that letter, i.e. it will not lead to the spacing which can rightfully be considered as the final spacing for that letter [why? see below]. (Note that a "spacing problem that is caused by the shape of a letter", is not a spacing problem at all — it is a problem with the design of the letter. Don't get fooled by sloppy language.)

    Because spacing and kerning should follow from glyph design, the spacing and kerning of a font can only be created after the design of all the glyphs of the font has been completed.

    So, although, during the phase of designing glyphs, spacing/kerning can be taken into account by a designer (see above) — the real spacing/kerning should be a completely separate step from designing glyphs. So I would agree with this: "Finally, done drawing all the characters. Now on to the spacing/kerning!"

    What cannot be separated, however, is the spacing and kerning, because they are interrelated.

    It is more difficult to create good spacing, than to create good kerning, because good spacing should take the complete kerning of the font involved into consideration. Good spacing results from an optimization process: it should minimize the (average) kerning that is needed for the font involved.

    To be able to create the kerning for a font, that font should have spacing (for all glyphs). To be able to create good/better/optimal spacing for a font, that font should have kerning (for all glyph pairs that need to be kerned; it's crucial to use a very good list of all potential kern pairs). Because of this circular or chicken-and-egg relationship, it is necessary to start with provisional spacing/kerning, because only in that way, it is possible to create, as a next step, optimal (or better) spacing/kerning. (Of course, the optimization process should have parameters which can be tuned by the font designer. The value of these parameters are the designer's input to determine the "style of spacing/kerning" — for instance, how narrow or wide the average spacing/kerning should be.)

    To be continued in the next comment...
  • Ben BlomBen Blom Posts: 248
    ... Continuation of the previous comment

    I would suggest that the application for auto-kerning and auto-spacing will be used to create the provisional spacing, and, after that, the provisional kerning, and, after that, the final spacing, and, after that, the final kerning. Because of the interrelatedness of spacing and kerning — an extra iteration after that, of creating spacing and kerning with the application, might improve the results. And of course, the designer can review the results and fine-tune them by eye.

    Fundamentally, creating kerning and spacing with an auto-kerning/auto-spacing application, can be regarded as an iterative optimization process.

    When designing an application for auto-kerning and auto-spacing, there are two main challenges: (1) design an algorithm to create good kerning for glyph pairs, and (2) design an algorithm to create optimal spacing based on the complete kerning of a font — and, of course, based on the design of each glyph involved. An additional challenge might be to design an algorithm (with parameters) to create good quality kerning classes; so the creation and use of good kerning classes can be integrated into the auto-kerning application, making it much more useful.

    I think that creating a good application for auto-kerning and auto-spacing is not very easy, so we should appreciate anyone trying to do so.
  • You're either a reallllllllllly good type designer if you think spacing should happen after you've finished drawing...

    ...or you're a not very good one.
  • James PuckettJames Puckett Posts: 1,699
    You're either a reallllllllllly good type designer if you think spacing should happen after you've finished drawing...

    ...or you're a not very good one.
    Are you mocking the creator of UltraPrecision™ font technology?
  • edited August 2012
    I was trying to allow for a level of clarity of thought, vision, and planning that probably exists at level of type design genius.
  • My 2¢: just to place oranges with oranges, and so on, if my letterforms are dictated by some sort of grid geometry, I'd agree with Ben's statement. As for the apples, on text faces, especially serifs, the spacing informs the shapes and vice-versa, Ben's approach would fail spectacularly?
  • Ben BlomBen Blom Posts: 248
    "the spacing informs the shapes and vice-versa"
    That sounds very interesting... but what does it mean? I am not sure. I try to interpret this idea.

    (1) The spacing informs the shapes, and (2) the shapes inform the spacing.

    First I try to interpret the second part: "the shapes inform the spacing". I would interpret this as something like: The shape of a glyph "determines" or influences the spacing and kerning which should go with that glyph. When we consider the shape of a glyph, we expect some kind of spacing and kerning for that glyph. When we, for example, design a "Gill Sans R", where the part at the bottom right is sticking out more than in a "Trade Gothic R", we expect that (1) the right margin of the "Gill Sans R" will be smaller than the right margin of the "Trade Gothic R", and that (2) the "Gill Sans R" will be the left glyph in more kern pairs than the "Trade Gothic R" and/or that, the average absolute value of the kern pairs of which the "Gill Sans R" will be the left glyph, will be bigger than the average absolute value of the kern pairs of which the "Trade Gothic R" will be the left glyph. Is this a correct interpretation of the second part?

    Now to the first part: "the spacing informs the shapes". How to interpret this? Does the right margin of a glyph "determine" or influence how the shape of the right side of that glyph should be? If I increase or decrease the right margin of a glyph, would that make me feel inclined to change the right side of the shape of that glyph? I simply don't see how (a change of) the width of the right margin of a glyph, "determines" or influences (a change of) the right side of the shape of that glyph.

    One more try. If a glyph is the left glyph in many kern pairs, and if we want, in general, to minimize the number of kern pairs in the font involved, then one could say that that number of (expected) kern pairs might influence a decision to change the right side of the glyph involved — to reduce that number of kern pairs. Could we, in that situation, say that that number of kern pairs in itself "determines" or influences how the shape of the right side of that glyph should be? No, because without us wanting to minimize the number kern pairs in the font involved, that number of kern pairs would not be relevant for the shape of the right side of the glyph. Or, is it self-evident, that we, in general, want to minimize the number of kern pairs? OK, let's say it is self-evident. Would we then really want to change the right side of the glyph involved, to minimize the number of kern pairs of which this glyph is the left glyph? I don't think so, because to really minimize the number of kern pairs involved, we might have to change to right side of the glyph until it gets very ugly. Is minimizing the number of expected kern pairs more important that the esthetic qualities (whatever they may be) of the glyph involved? I don't think so. I'm stuck again. I don't see how the number of expected kern pairs for a glyph, in itself "determines" or influences the shape of that glyph.

    So I can understand it when someone says that the shape of a glyph informs the spacing for that glyph. But I cannot understand the opposite, that the spacing of a glyph informs the shape of that glyph. I can only see a one-way relationship with this, not a two-way relationship.

    Would anyone please clarify what "the spacing informs the shapes" means?
  • Someone else can do this better, let's hope, but for now: you are drawing your new letters. It's a serif typeface. You have drawn an H and an O. How far should the serifs of H extend? You add h and o and look at HO Ho ho OH Oh oh and OO and HH and oo and hh. Can you finish these 4 glyphs without considering the spacing?

    You have now added T and A, with serifs, and t and a and o. Then you try TO To to TA Ta ta, as well as TT tt AA aa. You have to consider spacing and kerning to see how they fit and how the serifs work. Mix up, complicate hundred times, repeat.
  • Jan, I think that your description of the meaning of "the spacing informs the shapes" is in line with my description above of what I would call "the phase of designing glyphs". To quote myself:
    So, although, during the phase of designing glyphs, spacing/kerning can be taken into account by a designer (see above) — the real spacing/kerning should be a completely separate step from designing glyphs. So I would agree with this: "Finally, done drawing all the characters. Now on to the spacing/kerning!"
    See the "above" this quote is pointing to.

    We seem to agree that, when designing glyphs (shapes), spacing/kerning can be used as a means to see glyphs in the context of other glyphs, and to see how those glyphs function together. So, we seem to agree that spacing/kerning can be considered as something that assists in the process of designing glyphs.

    However, we seem to disagree about the status of the spacing/kerning (if any) which is used during, and which results from, this designing process. You seem to consider the use of spacing/kerning when designing shapes, as mandatory, especially when designing a (serif) text typeface. I think is it possible to design shapes without the use of spacing/kerning, although I feel it might be useful in many situations. You seem to consider the spacing/kerning which results from this designing process, as final, or perhaps nearly final. I consider the spacing/kerning which results from this designing process (if any) as provisional at most.

    Although we agree, to quite some extent, about the use of spacing/kerning when designing shapes, I guess that we see this designing process from a different perspective or paradigm.

    Perhaps, using the "Finally, done drawing..." sentence is confusing and inappropriate, because it can incorrectly be read as suggesting that during the drawing of characters, spacing/kerning will not be taken into account by the designer involved. So it should be rephrased as something like this:
    "Finally, done drawing all the characters (and using spacing/kerning to assist in this drawing whenever needed). Now on to the spacing/kerning!"
  • Saying that spacing "can be considered" when designing glyphs is like saying that the principles of aerodynamics "can be considered" when designing an airplane's wings. The negative space within each letter is only correct or incorrect in relation to the negative space between letters; the only way to judge it is to set sidebearings and examine the letters in different combinations. The only way to judge the rhythm, color, and readability of a typeface is to view it as lines and paragraphs; you can't do that without spacing it. Obviously the spacing won't be 100% final until the typeface is done, but the glyphs and the space around the glyphs have to be designed together or the final result won't be balanced.
  • Max, I'm afraid you misread me. I did not say or imply that spacing "can be considered" when designing glyphs, as something that is irrelevant for the design of the glyphs. I said that spacing/kerning can be considered as something that assists in the process of designing glyphs — which is quite something else.

    Designing an airplane's wings, without applying the principles of aerodynamics, might lead to a disaster. Will designing the shape of, let's say, an "O", without applying the principles of spacing (whatever they may be), always lead to problems? Will it necessarily lead to an unbalanced font? I just don't believe this. The same for designing an "H". Etc. Of course, after those glyphs are designed, the final spacing and kerning has to be added. And of course, during the design of a glyph, combining the glyph with other glyphs and using provisional spacing, can be used to judge the design of the glyph involved in the context of other glyphs, and this might lead to changing the design of that glyph.

    The comparison with the airplane's wings and applying the principles of aerodynamics, suggests that there are general principles of spacing. Are there really universal principles of spacing which we have to know, and which we have to follow deliberately? (Universal principles which are followed "automatically", are not relevant here, because we don't have to know them to be able to use them. When riding a bicycle, I might follow many universal principles I am not aware of — but I don't have to know these principles to be able to drive a bike normally.) The only "universal principles" I know of, when spacing a font, are my eyes.

    "The only way to judge the rhythm, color, and readability of a typeface is to view it as lines and paragraphs; you can't do that without spacing it." Of course, but from this it doesn't follow that "the glyphs and the space around the glyphs have to be designed together or the final result won't be balanced". Are the fonts of Jos Buivenga, which have been auto-spaced and auto-kerned by Igino Marini's iKern (http://ikern.com/k1/category/collaboration/page/3/), imbalanced? Is Brandon Grotesque by Hannes von Döhren, which has also been auto-spaced and auto-kerned by iKern, imbalanced? Are the fonts of the Dutch Type Library, which have been auto-kerned by KernMaster (http://www.fontmaster.nl/), imbalanced? Can anyone please give an example of such an imbalance?
  • Depend on it, Ben—Buivenga, von Döhren, and the folks at DTL space all their fonts carefully before turning them over to iKern or KernMaster. Marini's program supplants the original spacing with its own—he's told me he can't, as the program is currently written, kern a face without respacing it—but he makes every effort to stay close to the designer's original intention, and each face is reviewed and annotated by the designer several times before the new spacing and kerning are finalized.

    And yes, there are universal principles of letterspacing. I mentioned the principal one: a letter's internal spaces must be in balance with the spaces to either side of it. But there's also what could be called the please-keep-your-hands-inside-the-moving-vehicle-at-all-times rule, which states that if any bit of your glyph pokes out too far to the left or right, you will probably be sorry. For an example of someone ignoring this rule and designing a pretty letter that cannot be satisfactorily combined with others, see the leg of the R in Bembo and Trajan.


  • Matthew ButterickMatthew Butterick Posts: 143
    edited September 2012
    But I cannot understand the opposite, that the spacing of a glyph informs the shape of that glyph.
    The simplest example would be a monospaced family.

    A less simple example would be styles that are required to match the metrics of each other, either for functional or aesthetic reasons.
  • Type designers are conservative.
Sign In or Register to comment.