A new name and some tweaks to the roman

Even though I’m working on italics, there were a few things I wanted to fix in the roman style (a type designer, or any artist for that matter, should never be afraid to go back and add some polish to previously completed parts of the font). Here, I made the letterforms a bit wider and rounded the bowls some.

The difference isn’t big, but I think it makes it slightly easier to read, especially at medium sizes (11–16 pt). (Text is from Wikipedia)

The one on the right is the new version of the font.

In other news, I’ve also decided to change the working name of the font. Instead of “Floribunda”, the typeface will now be known as “Inflorescence” instead. I think it sounds a bit nicer (no forward-schwas in the middle of the word!).

Introduction to italics, part one

At this point, we’ve completed the entire ASCII block (save for the at sign (@) ).

Now, we could go on to do one of four things. We could add more glyphs to the font—fill out the Latin Supplement block with accented letters like ‘ å ’ and a few moderately common symbols like ‘ ¢ ’ and ‘ ° ’. We could go back and actually kern the letters we’ve already made to make the spacing look nicer. We could also derive the bold weight, or we could create the italic version of the font.

I recommend doing italics first. That is because between these four projects, creating italics will add the most value to your font. A functional font family includes a regular, or roman style, and an italic style. In many cases, italics are mandatory, for setting names of works of art, for example. They are also the best way to emphasize words in text. What about deriving the bold weight, you may ask? Well, bold is much, much less commonly used than italics. A font family does not need a bold weight to be functional, though it can be a selling point. Bold is usually used in contexts where it would also be appropriate to use a completely different typeface (headings, for example). The bold weight is also much more straightforward though tedious to design. It is also much easier to embolden both a roman and an italic typeface to create the bold and the bold italic than to italicize both a regular and a bold font to make the italic and bold italic. That is why I like designing italics before bold.

What are italics?

Italics are much more different from the regular style than the bold style is. In serif type, it is possible to machine-produce a passable, though ugly, bold font given a regular font, since emboldening is a relatively straightforward and mechanical operation. Indeed, dozens of weights can be automatically interpolated (with quite good results) from just three base fonts (extra-light, regular, and black). However it is impossible for a computer to automatically produce the italic style of a serif font (something the opposite is true in sans serif type, where the “italic” is very close to a machine slanted version of the original, and the different weights are given much more consideration). That is because the serif italic letterforms themselves are drastically different from the roman letterforms. Italic letterforms are much closer to calligraphy than roman letterforms are. In fact, the italic hand is one of the most natural styles of calligraphy. This is evident in the strokes of the italic letterforms, which are more cursive than those of roman letterforms.

(Somewhat messy) italic calligraphy of some lyrics from How You Get The Girl, in Taylor Swift’s 1989. While handwritten, this calligraphy is essentially the same as italic type. Compare this with the Proforma Italic sample below.

For comparison, here is a sample of the italics from the Proforma font (grouped by letter shape):

The letterforms are much more free-flowing and give a completely different texture than the roman style does. Here are some important differences.

Testing your font in print

We’ve made test texts before, but those were all at the sentence or the paragraph (at a stretch) level. Now that I have a full upper and lowercase glyph set, spaced and all, and some rudimentary punctuation marks, I can set entire passages of text in Floribunda. (Since the last post, I’ve spaced the capitals and added some basic punctuation marks— ‘.’ ‘-’ and ‘,’ . Spacing capitals is done basically the same way that the lowercase letters are spaced; punctuation will be covered in a later post).

Producing test pages is not hard. Just grab some text from a random Wikipedia article (preferably one that contains few numbers or special characters), and format it into columns set in your font. Throw in different-point-sized and ALL CAPS text generously—the capital text in mine is set in faux small capitals (the most bearable of the four typographical ‘faux’s, in my opinion anyway)—and you might want to include some knockout and grayscale type as well. Then just print it out on paper (please be kind to the environment and reuse some computer paper that’s been printed on on the other side—type design does not require blank new paper). Here is the page I used; the passage is taken from the Wikipedia article on Auroras.

Note that the ideal body text font size to use is much smaller than you’re used to specifying. For most typesetters, 12 point type is clumsily gargantuan, and generally reserved for headings and other display text. Most body text in magazines and textbooks is set in about an 8–9 point font (this corresponds to about a 10–11-pixel em square). Don’t believe me? Print out some nine point text and compare it to a copy of TIME magazine. The type will probably be about the same size.

Strategies for setting letter spacing: part two

In the last post, we resolved the sidebearings of three letters—‘n’, ‘o’, and ‘v’, each representative of a class of letter shapes. These sidebearings can now be assigned to related characters with congruent sides. The sidebearings of the ‘n’ also work for the ‘m’ and the ‘i’ (but not necessarily letters like ‘l’ which has a taller stem and no head serif on the mean line). Since the left side of the ‘n’ is the same as the left side of the ‘r’, the left sidebearing of the ‘r’ can be taken from the corresponding sidebearing on the ‘n’. The same goes for the right side of the ‘h’, which matches the right side of the ‘n’. In Floribunda the right side of the ‘a’ also matches the ‘n’, but this is not always true.

The sidebearings on the ‘o’ can inform the ‘e’ and ‘c’ (the right sidebearings might need tweaking) and the bowls of the ‘b’, ‘p’, ‘q’, and ‘d’. The ‘v’s sidebearings set the sidebearings of the ‘w’ and the ‘y’.

Strategies for setting letter spacing: part one

At this point, all twenty-six lowercase letters have been drawn and packaged into a font file, but they lack any horizontal metrics, so they all print in their own em squares if you try to type with it.

Text credit: Wikipedia: Sun

We have to set the sidebearings (basically the width) of each letter to get the text to print more like this (again, capital letters and punctuation have been stripped since Floribunda doesn’t contain them yet) :

Floribunda has a complete lowercase alphabet!

Here’s a cute pangram (sentence that contains every letter of the alphabet) typeset in Floribunda (forgive the missing commas!) :

The font file contains no side bearings or kerning—each letter is in a 1 em square, so it’s just if you want to check out the outlines or something. The pangram was typeset with individual letters. Download it here.

Glyph design: the lowercase z

o
i
a

b
l
n

d p q
e
f
v
h u m
r

c
j
t
y
w
x

k
z

g
s

Bringing up the end of the lowercase English alphabet, and the lowercase bloc of Floribunda, is the letter ‘z’. The ‘z’ is an incredibly rare letter, the rarest English letter in fact. On average, it only occurs once every 1,350 letters (‘q’ occurs once every thousand letters, and ‘x’ every seven hundred). But while ‘z’, like ‘x’, is rare, it is vastly important because it bridges the gap between the lowercase and uppercase alphabet.

The ‘z’ is unlike any other lowercase letter. It defies the rules of calligraphy—based on the stroke angle, all parts of the ‘z’ should be hairline. But since you can’t have a letter made completely out of hairlines, it’s become a convention to embolden the diagonal (the convention is reversed in italic type). The letter’s two horizontal arms are the most important parts of the letter. No other lowercase letter contains straight horizontal strokes (besides unserifed crossbars like in ‘t’)—the miniscule alphabet always converts such shapes to bowls or arches like in ‘f’ or ‘p’. But these strokes are common in the uppercase alphabet, making ‘z’ a useful base to start constructing letters like ‘E’ or ‘L’. They will also be extremely handy in constructing lowercase cyrillic letters, if we ever get to that.

Components of a lowercase ‘z’

The ‘z’ is not hard to construct. You can get the diagonal of the ‘z’ by reflecting the ‘x’ horizontally. Normally you should never ever reflect a letter in a single axis as it reverses the stroke stress, but we can get away with it with the ‘z’ since its diagonal is already a stress violation anyway. The diagonal is the same length and width as the one in the ‘x’—but the ‘z’ is slightly narrower because it lacks the serifs that come off the diagonals.

How to deal with font extremas

I don’t know about you, but this is definitely my least favorite font error. If your glyph outlines aren’t constructed in just the right way, fontforge will complain about missing “extremas”. But what are extremas anyway?

An extrema is a special kind of point that helps define the edges of a glyph. The word is Latin—technically a single extrema is called an extremum, with extrema being the plural. But the term has become so common among type designers, that extrema and extremas have become the terms used to describe these points in a glyph outline.

An extrema is just a curve’s minimum or maximum in the xy coordinate plane. This is just like finding the maximum of a parabola or some other curve, except applied to bézier curves.

Each bézier segment of a glyph’s outline (read: each segment between two points, not the entire glyph outline) must be a one-to-one function*—i.e. it has to pass both the horizontal and vertical line tests. A bézier segment in a font shouldn’t change direction in either the x or the y direction—otherwise it should be split into two one-to-one segments. A handy (but not foolproof) way of checking this is looking to see if the curve fits into the rectangle determined by its endpoints (though an ‘s’ shaped curve can pass this test and still not be one-to-one).

The curve on the left is acceptable—it doesn’t change direction in the x or y axis. The curve in the middle folds back on itself, so it needs to be split in two (right).

Why is this a thing? Well apparently it helps with hinting, by letting the renderer know the bounding boxes of the glyph outline. It also makes it easier to measure stems and set sidebearings, since you can measure distances between points instead of vague spots along a curve. But the truth is, most renderers get along fine without any extremas. No vector illustration program requires extremas on its paths. Extremas are just extra points on the ends that help protect glyph outlines, since fonts undergo much more heavy-duty scaling and rasterizing than most other forms of vector art. Think of them as protective bumpers on your letters that keep your font from getting distorted by dumb renderers.

The forest and the trees: How to test your font

As I’ve said before, it’s important to periodically combine your glyphs into words and see how they work together. You should already be examining your typeface at the word level. At this point, Floribunda has enough letters that we can also start assembling short test texts. Test texts can be as long as a paragraph or even a page, but for now, we’re just looking for problems at the line level. A few issues I saw right off the bat was that the ‘f’ terminal was a bit too light, and the spur of the ‘a’ too thin.

A test text for Floribunda. Test texts don’t necessarily have to make sense,
they just simulate font usage at the line, paragraph or page level.

I also wasn’t a fan of the bowls on the ‘b’ and derived letters, so I redrew them with fresh curves.