How do you maintain letter-form consistency?

Type designers always say "trust your eyes", but in fact, once you start working on complex type-design projects (multiple masters, large glyph sets, etc.), you realize this nonsystematic approach has its limits. At the very least, it's time consuming.

That raises the question of how to maintain letter-form consistency when simple copy-paste isn't good enough.

For instance, making the top and bottom halves of an s look "the same" is hard and time consuming. I've recently found myself involving all sorts of measurements in the process. Another example is equalizing the apparent tension (curvature) of rounded shapes which are related but not identical.

How do you approach this issue? Do you freely move nodes around while relying solely on your eyes? And if not, what do you do? If you could describe some key examples in detail, that would be awesome.

On a related note, I'm also interested in what compromises you make in order to keep things simple and manageable. I mean stuff like "all my overshooting letters overshoot by the same amount regardless of their shape", "I skip making letters such as I,l a bit wider", etc.



  • I'm quite a beginner, but I hope I can give my opinion.
    About overshoot, I just make the same amount, usually 8-16 unit on 1000 em, depending on style or width.

    About curve tension, I'm using Fontlab 7, so this feature comes built in. Press Ctrl + Alt + L turns on the Tunni lines and we can see the tension, usually I try to keep the numbers consistent.

    About the overall consistency, I try to use longer text in 24pt or less to scan through the text and read them, then zooms in on letters that jumps out.

    Would love to hear others' opinion on this too.
  • To me the challenging question is: how do you maintain inconsistency.  :-)  Because the grid*, copy-paste, and symmetry are handy tools of consistency. Overly handy... Because they end up betraying the eye. So the key is knowing when to let go of certainty. Not least to save time! To me that is Design, and possibly even Life.

    So for example in an "S" (with contrast) I would start with a spine exactly as thick as the diagonal of the "N", make it thicker by some amount (which I record for possible re-use) because it's a curve and not a line, and place it higher than the middle by some amount (centering to the bar of the "H", although crucially only provisionally). Depending on the desired relative widths of the "S" and the "N" I would rotate the spine (BTW probably not bother recording that number). Then I would move to the top and bottom, overshooting them by the predetermined amount, and making them diverge in width depending on the desired mood of the design. If there are serifs I would start by copying the one in the "G", but change their sizes (by some, probably recorded amount) depending on the relative counter sizes. And then I let go. The rest is beyond conscious grasp (although any amounts recorded above that shift do get noted for possible re-use elsewhere) and the eye takes over, jumping between the various parts of the letterform, iterating until it's essentially no longer *reasonable* to do so. BTW only your eye, and only at that moment – it will for sure shift over time, although with experience less than an –actionable– amount.

    So to me it's essentially a hybrid between the exact and the unknown (which I argue is furthermore *unknowable* in the end). Crucially, it's a non-deterministic process, where certainty doesn't get you nearly as far as one might like... Denying that yields poor results, such as unmodulated calligraphic/geometric fonts.
  • automatically make it thicker by some amount
    An elaboration: the "letting go" there can happen early or later, the latter by starting with the same amount you diverged the thickness of the curves from the straights by. But you will probably have to diverge from *that*, so I prefer the former; it saves time, with no –knowable– loss.
  • Chris LozosChris Lozos Posts: 1,456
    nonsystematic approach has its limits. At the very least, it's time consuming.
    Yes, but the converse is also true. This may vary with the skill set of the user, and their degree of faith in either mathematical calculation or visual acuity. It is my experience that deviation from the precise "number" of a calculation is not cheating.  As rational beings, we may suppose that we are in awe enough of tidy numbers to assume they ensure accuracy. But if we define accuracy as only matching numbers, we miss the point that there are no rulers affixed to every glyph in normal viewing. We "see" relationships that our senses present to our determining brain. This ability to "measure" is relative, not absolute. We can sometimes see when something looks out of character, even if we don't know why. Overshoot is a perfect example. We all accept the need for it, even if it muddies up our calculations. The human being is gifted with the ability to "see" relationships and deviations [perhaps a product of being able to distinguish prey from predator in early humankind]. Consider this: When balancing a tree branch over a fulcrum, is it easier to calculate the weight variation and length mathematically then mark the balance point at one take or is it easier to just place the branch on the fulcrum and slide it until balanced? (If the the branch is a perfect cylinder like a dowel or if it is irregular like a true branch in nature.) An example might be balancing a Helvetica H vs a Bodoni bold italic ampersand?
    What I think is that people want to be SURE they are correct and can prove it with a measure. They fear being contested in their positioning because the numbers are too fuzzy to be convincing. We live in a "fuzzy" World that Science may at some future point make less fuzzy.  Till then, feel free to either use your eyes or spend hours doing calculations, whichever pleases you.

  • I agree with everything here. I trust my eyes more than anything, but starting with the same measurements is often the right approach. Don’t worry so much about measurements if it’s keeping your design from looking right.

    Also, it’s less important than it was 30 years ago, but keeping matching numbers (stem widths, overshoots) the same is important when there’s no reason for them to be different. It’s very common for, say, a lowercase stem to drift off by a unit or two for no good reason. Making sure numbers are the same where they are intended to be means more consistent auto-hinting, rendering, etc.

    (And understanding how and when hinting works, and what the tolerances for error are, is helpful when one starts make adjustments by eye.)
  • Ori Ben-DorOri Ben-Dor Posts: 381
    Thank you all for some very interesting answers!

    @Hrant H. Papazian, when you say you record measures, what does it look like in practice and what do you do with that later? Are you like adding a note to the glyph saying something like "for the serif I took the one from G and scaled it 85%"? And then later when you draw another weight you'd use that to repeat the the same process?
  • Yes, although I usually deal with em units not percentages (to avoid rounding errors). BTW not just across styles, other glyphs within a style can borrow the measure too. And any measure that gets re-used enough ends up promoted to the global "Font note" area.
  • AbiRasheedAbiRasheed Posts: 236
    edited April 2021
    I've used FLV's (havent played with FL7 yet) tunni lines for curve tension which are super useful, so is harmonize and the rapid tool. Aside from that I usually rotate it to see what it looks like from a different angle, more often than not you'll see tension that is more or less than that is desired. Rotating it will give you different curve tensions in different angles, so sometimes you bet on the eye to do the guess work. End of the day the typeface is front facing so whatever curve tension is pleasing to look at at that angle is the winner irrespective of how odd it looks like say upside down or side facing etc. 
  • Adam JagoszAdam Jagosz Posts: 688
    edited April 2021
    I think John Hudson’s comment from here is relevant:
    Understand the difference between optical deviations within a system and arbitrary measurements. The more things you make quantifiably systematic, the easier the work will be, especially when you start expanding your design into multiple weights or widths, size-specific instances, extending kerning, etc.. The more measurements—stems, bowls, counters, sidebearings, kerning values—that are arbitrarily variant, the more difficult all these taks will be, because you will be constantly needing to take the variance into account. So, yes, optically deviate from mathematical standardisation when necessary, but start with the standardisation and track the deviations. Most crucially, if you deviate in a particular way in one part of the design, deviate in the same way in all related parts of the design: don’t apply different methods arbitrarily across a design space.
    In short, keep the inconsistencies consistent.
  • Ori Ben-DorOri Ben-Dor Posts: 381
    Can someone please give a real-life example of how you record all those measurements that deviate from mathematical standardization? I understand the concept in theory, but when I imagine myself starting adding notes to glyphs with that information, I see a lot of mess...
  • Well I write it on paper first, and once I settle on the number I note it in the font. But there really isn't a ton of it, I personally only record things that I think I'll probably re-use. And if I end up needing a number I didn't record I can always go back and re-measure.
  • Chris LozosChris Lozos Posts: 1,456
    I just copy into the background the routine I wish to follow and adjust as needed.
Sign In or Register to comment.