Skip navigation
Help

Alexander Charchar

warning: Creating default object from empty value in /var/www/vhosts/sayforward.com/subdomains/recorder/httpdocs/modules/taxonomy/taxonomy.pages.inc on line 33.

  

What is it that makes a typeface into a text font, instead of a font for larger sizes? The answer differs slightly, depending on whether one aims for print or Web-based environments.

Nevertheless, there are certain features that most good text faces have in common. Familiarity with these helps to select the right fonts for a given project. This article presents a few criteria to help the process along.

Some of today’s most successful typefaces were designed to excel in very specific areas of use: Frutiger grew out of airport signage, Georgia and Verdana were among the first mass-market fonts created for on screen reading, FF Meta was conceived as a telephone book face, and even the Stalwart Times New Roman was tailored for the pages of the London Newspaper The Times. Many typefaces are also often fine-tuned for using in certain sizes.

It should be noted that in this article, when “text” is mentioned, it is in discussion of body text, or running text (in other words, text at a similar size to what you are probably reading right now, rather than much larger sized words).

Features Of A Good Text Typeface

The features outlined in this article are those that type designers keep in mind while developing new typefaces. It’s important to realize that these aspects of typeface design are different from the text treatment a graphic designer employs while laying out a book page or website—no matter what a typeface’s inherent rhythm and niceties are, setting a text is still something that must be done with great care in respect to readability. There are problems that good fonts themselves cannot solve—whether or not a text sings on the page or screen depends on factors like the width of the column, the amount of space between each line, the contrast between the foreground/background and a number of other factors.

Different versions of the Bembo design
Above, Bembo over the years: this typeface was a favorite of many book designers throughout the 20th century. At the top of the image is a scan of the original Bembo typeface, printed with letterpress. The digital version of the typeface—Bembo, seen in the middle, is too light for ideal text in print. A newer digitization was published in 2002—Bembo Book, seen at the bottom. This font is much darker, and is a better representation of the original Bembo idea. However, the middle version is still very elegant, and may still be used well in sub-headlines.

Every typeface has its own inherent rhythm, created by the designer who made the font. With typefaces that are intended for use in body text, it is primarily this rhythm that will make the typeface readable. But there are additional factors that go into the making of a good text face: the space between the letters, the degree of contrast in the letters’ strokes, as well as the x-height and relative size of the whitespace inside of the letters. Not every typeface that works well in text will apply all of these factors in the same way, but all good ones will have many of these features in common.

1. Stroke Contrast

When it comes to typefaces, the term “monolinear” is used to describe letters that appear to be designed with a consistent stroke thickness. Monolinear typefaces are low-contrast typefaces. Stroke contrast can be a helpful feature in small text sizes, but it is not paramount that a text face appears to be monolinear. Indeed, many newspapers employ high-contrast fonts; the question that must be considered is just how thick the thin strokes in high-contrast typefaces are.

Sample Layout in the Cycles typefaces

The images in this section show different ends of the contrast spectrum: the Cycles types shown above are serifed, with a good deal of contrast. Sumner Stone’s Cycles typeface is an excellent choice for book design as its letter forms combine clarity with a rather high degree of stroke contrast and an almost timeless appearance. Five separate “versions” of Cycles are used in the above image; each block of text is set in its own optically-sized font.

Below, Avenir Next—also a great text face—is from another style of letter, and has very little contrast. I wouldn’t split good typefaces up into good contrast and bad contrast groups. Rather, some typefaces have a degree of contrast—be it too high or too low—that makes them less suitable for use in text. There is no definite rule on how much or how little contrast impacts a text face’s legibility. However, it is clear that both no contrast and excessive contrast can have adverse effects.

Text in Planeta and Avenir Next

Geometric sans serif typefaces often appear to be monolinear stokes; their letters seem not to have any stroke contrast. In order to achieve this effect to the max, type designers have always made slight optical corrections. To look monolinear, a geometric sans needs some degree of thinning. In the image above, Planeta (left) is compared with Avenir Next (right). Both typefaces are more recent additions to the geometric sans category than stalwart faces (like Futura), or classic display designs (like ITC Avant Garde Gothic). Planeta has no visible stroke contrast, which must be a conscious decision on the part of its designer. While this does give it a unique style, it makes the face less suitable for text than Avenir Next, which is actually not as monolinear as it appears at first glance.

2. Optical Sizes

Text in Garamond Premier Caption and Display Sizes

The Garamond Premier typeface family features different versions of each font. These variants are tailored for use in a certain size range. Above, the Display font (left) is compared with the Caption font (right). The Display font is optimized for texts that will appear in very large point sizes, while the Caption font has been optimized for very small text.

In her book Thinking with Type, Ellen Lupton writes:

“A type family with optical sizes has different styles for different sizes of output. The graphic designer selects a style based on context. Optical sizes designed for headlines or display tend to have delicate, lyrical forms, while styles created for text and captions are built with heavier strokes.”

The intended size of a text should be considered when selecting the typeface: is the typeface you want to use appropriate for the size in which you need to set it? Does the family include optical sizes (that is, different versions of the typeface that are tailored specifically for use at different sizes)? As with each of the factors mentioned in this article, the size at which a font is set can make or break your text.

In many ways, it is easiest to see the qualities necessary for good text faces by comparing potential selections with “display” faces. Like the term “text,” “display” refers to the size at which a specific font may best be used. In print media, as well as in many screen and mobile-based applications, the term “display” is often analogous with “headlines.” If a typeface that you are considering looks more like something that you might like to use for a headline, it won’t be the best choice for body text.

In the comparison image below, the Garamond Premier Display font has a tighter rhythm than the Caption font—not as much space is necessary between letters when they are set in large point sizes. Why should one consider type families with optical sizes, anyway? Well, as users bump up the point size of digital fonts, the space between letters increases in equal proportion. This inter-letter space slowly becomes too large, and makes a text feel like it is breaking apart. When a proper text font is set large, it may require some tighter tracking. Typeface families that offer optically-sized variants of their styles play a helpful role here.

Text in Garamond Premier Caption and Display Sizes

In the image above, the first line of text—“Stanley Morison”—is set in the Garamond Premier Display font, while the lines of text underneath it are set in Garamond Premier Caption. Each font is balanced for its size, and they also harmonize well with one another. In another image (below), these fonts have been switched: the headline is now set in the Garamond Premier Caption font, and the text in the Garamond Premier Display. The letters in the Caption face look too clumsy when they are set so large, while the Display fonts’ letters appear uncomfortably thin in a “text” setting.

Text in Garamond Premier Caption and Display Sizes

The amount of stroke contrast visible in caption-sized fonts is much lower than in display-sized fonts. If the Garamond Premier Display font (from the above image) was rendered in a smaller point size, its thin strokes would begin to break apart, making the text unreadable. But this would not occur with the Caption version.

Garamond Premier Caption can robustly set real text, even in poor printing conditions. How well a font will render in small sizes on screen depends on the operating system and applications in question. Font formats themselves also play a role; in certain environments, TrueType fonts with “hinting” information may vastly improve on screen display (see the “Hinting” section at the end of this article).

3. x-Height

Text in Garamond Premier Caption and Display Sizes

Garamond Premier’s Display face (above left) is shown next to the Caption face (above right). Both fonts are set at the same point size. The Caption face features a much higher x-height than the Display font.

Many successful text faces feature high x-heights; this means that the ratio of the central vertical area of lowercase letters—the height of the letter x, for instance—is large when compared to the length of the ascenders and descenders. Depending on its design, a text face may have a low x-height and still be quite legible. But the benefit of incorporating a large x-height in a design is that it maximizes the area of primary activity.

A high x-height may also prevent some letters, like the a or the s, from appearing to become too dark; these two letters have three horizontal strokes inside the x-height space, which is a very small area in text sizes. In order for letters to maintain clarity and understandability, they must have a consistent rhythm, as well as include large, open forms.

4. The Spaces Inside of Letters

The images below illustrate just a few of the intra-letter spacing elements that should be understood and considered when choosing which typeface to choose for your body text. In order for the white spaces inside of letters to remain visible in small sizes, it is necessary for their counterforms to have a certain minimum mass, proportionally.

Counters
ITC Bodoni Six and ITC Bodoni Seventytwo

The image above shows text set in two members of the ITC Bodoni family: ITC Bodoni Seventytwo and ITC Bodoni Six typefaces. In the first line, “Randgloves” is set in a size mastered for 72pt display (ITC Bodoni Seventytwo), while “and jam” is in the Caption size (ITC Bodoni Six). These words are reversed in the second line. Note how the enclosed white space in the top portion of the e changes between the display and text optical sizes.

Apertures
Apertures in FF Meta

“Apertures” are the gateways that whitespaces use to move in and out of the counterforms of a typeface’s letter. The above image highlights the wide apertures in four letters from Erik Spiekermann’s FF Meta typeface. These allow for the typeface’s letterforms to feel more open. In certain sizes and settings, wide apertures—and the large counterforms that are their result—will make a text more readable.

Apertures in Frutiger and Helvetica

The top line of the image above is set in Helvetica, and the bottom line in Frutiger. While the counterforms inside the letters of these two typefaces are similar in size, Helvetica’s apertures are much smaller. Because of this, white spaces inside of Helvetica’s letters and between Helvetica’s letters are much more closed off from each other than in a typeface with more open counters—like Frutiger.

Other counterforms and problematic letters worth remembering include the c; if the apertures of a, e, s are very open, the c should follow this same route. Then there are lowercase letters like a, e, g, s that often have rather complex shapes—specifically, they each feature several horizontal strokes inside a small amount of vertical space. How do their forms relate to one another? How large is the typeface’s x-height? Do the ascenders and descenders have enough room, particularly f and g? Do the counterforms inside of roundish letters (e.g., b, d, p, q, o) have the same optical size and color as those inside of straight-sided letters like, h, n, m, and u? How different from one another are the forms of the capital I, the lowercase i and l, and the figure 1? Can the 3 and the 8 be quickly differentiated from each other? How about the 5 and the 6?

5. Kerning

Sample text in Carter Sans, with and without kerning

In the sample above, kerning has been deactivated for the second line. The gaps between the letters T y and V o are too large when compared with the amount of space between the other letters in the text. The typeface used in the image is Carter Sans.

Despite the popular misuse of the term in graphic design circles, “kerning” does not refer to the spacing values to the left and right of the letters in a font. Rather, fonts contain a list of kerning pairs to improve the spacing between the most troubling lettering combinations. The importance of kerning in a font is the role it may play in maintaining an optimal rhythm. Just as kerning describes something much more specific than a typeface’s overall spacing—or the tracking that a graphic designer might apply to a text—kerning is not the rhythm of a typeface itself, but an element that may strengthen a typeface’s already existing rhythm. Not every typeface design requires kerning, and there are typefaces on the market that indeed may have too many kerning pairs—a sign that the basic letter spacing in the font could have been too faulty in the beginning.

6. Consistent Rhythm Along the Line

Simple Text Sample in Frutiger and Helvetica

In the image above, compare the spaces between the letters of the Helvetica typeface (first row) with Frutiger’s (second row). Frutiger is a more humanist design, featuring a slight diagonal axis in its letters; many of them look similar to Helvetica’s, at least at face value. However, the space between Helvetica’s letters is much tighter.

While most of the images in this article feature typeface families that include Optical Size variants, many commonly used typefaces on the market today do not offer these options. This is why it is helpful to be able to identify text typefaces based on their features, rather than just on their names in the font menu. As mentioned earlier, it is primarily the typeface’s rhythm that dictates the readability of a block of text.

Take Frutiger and Helvetica, which are both commonly used in text, especially for corporate communication—Neue Helvetica is even the UI typeface in iOS and MacOS X 10.7. Yet, despite its popularity, Helvetica is not very effective as a text typeface; its rhythm is too tight. By rhythm, I’m not referring to tracking—or any other feature that a designer can employ when typesetting—but the natural flow of space between letters, and within them as well. Frutiger is a much more open typeface—the spaces between its letters are closer in size to the white spaces inside of the letters than in the case of Helvetica. Like all good text typefaces, Frutiger has an even rhythm—space weaves in and out of the letters easily.

7. Caveat: Signage Faces

To round off my discussion on text typefaces, I’d like to briefly mention some fonts that are often shown in rather large sizes: fonts for signage. Interestingly, many signage typefaces have design features very similar to typefaces created for very small applications. The Frutiger typeface, based on letters that Adrian Frutiger originally developed for the Roissy airport in Paris (now named after Charles De Gaulle), is quite legible in small sizes precisely because it is a good signage typeface. Despite their size, signage fonts serve a rather different purpose than Display fonts.

Frutiger in an airport signage-like setting

Additional Elements To Consider

After considering the criteria mentioned above, the next question that often comes up is, “does this font have oldstyle figures, or small caps and ligatures, etc.?” A font’s letters might look really great in text, but if they do not include additional elements and features, their use is somewhat minimized. I avoid using fonts with small character and feature sets where I can, because I feel that the lack of these “extras” may break the kind of rhythm I aim to achieve.

1. OpenType Features

Once you’ve established a consistent rhythm by setting your text according to the correct size and application, it would be a pity to inadvertently break that flow. Large blocks of tall figures or capital letter combinations do just that.

Even in languages like German, where capital letters appear at the start of many words, the majority of letters in a text planned for immersive reading will be lowercase letters. Every language has its own frequency concerning the ratio of “simple” lowercase letters like a c e m n o r s u v w x z to lowercase letters with ascenders or descenders—b d f g h j k l p q y. In international communication, language support is a key consideration when choosing a font, and other character set considerations may especially play a role.

FF Meta Pro Book and two examples from its many figure styles

Traditionally, the style of figures used in running text also have ascenders and descenders. These figures—often called oldstyle figures or text figures—harmonize better with text than the “uppercase” lining figures. These so-called lining figures either align with the height of a typeface’s capital letters, or are slightly shorter. It is no surprise that, when shipping the Georgia fonts for use onscreen and online, Matthew Carter and Microsoft made the figures take the oldstyle form. Many other typefaces that have long been popular with graphic designers, like FF Meta (seen above), also use oldstyle figures as the default style. In my opinion, lining figures are best relegated to text set in all-caps.

Long all-caps acronyms—like NAFTA, NATO, or USSR—also create an uncomfortable block in the line for the reader. Setting these letter-strings in small caps helps reestablish a specific typeface’s natural rhythm in reading sizes, as may be seen in the first line of the image below (set in Erik Spiekermann’s FF Meta).

FF Meta Pro Book and its small caps

Along with common ligatures like fi ff fl, small caps and the many figure options are the most common OpenType features found in quality text fonts. Aside from having both lining and oldstyle figures, OpenType-functionality can enable a font to include both tabular and proportionally-spaced figures, numerators and denominators for fractions, as well as superior and inferior figures for academic setting. Additional OpenType features (such as contextual alternates or discretionary ligatures), are more powerfully noticed in display sizes, and in some cases can even be distracting in text.

2. Hinting

The display of text on screen, particularly on computers running a version of the Windows operating system, may be fine-tuned and improved with the help of size-specific instructions inside of the font file. These instructions are commonly referred to as “hints.” A TrueType font (or a TrueType-flavored OpenType font), is capable of including hinting. However, not every font manufacturer goes to the effort of optimizing the onscreen appearance of its fonts for Windows—even those fonts specially created for use in text sizes.

Prensa in three different rendering environments

All of the text in the above image is shown in the same font: Prensa, set at 18 pixels. The lowest row shows this at actual size in three different onscreen rendering environments. In the enlargements, the top row shows a close-up of rendering in Safari on MacOS X, which ignores the hinting data in fonts. The second row shows rendering in Internet Explorer/WindowsXP (Grayscale only, for this sample). The third row is from a ClearType environment—in this case, from Firefox on Windows7. Prensa is a typeface designed by Cyrus Highsmith at the Font Bureau; the Web font is served by the Webtype service.

Recommended Typefaces For Readability

Aside from the typefaces already mentioned in this article and its images, here is a small selection of faces that I personally enjoy at the moment. Even though lists of “favorite” typefaces are about as useful as lists of favorite songs or favorite colors, I am happy to pass my subjective recommendations along. No doubt that as new projects arise, my list of favorites is likely to change, too. I do think that these typefaces serve as great starting places. Some are also just from cool friends whose work I dig. Alongside each selection, I mention whether this choice is currently available for print only, or if there is a Web font version, as well. Don’t forget: the typefaces that you pick in the end should depend on your projects, their audience, and the content at hand.

Small sample of the Arnhem typeface

Arnhem is a no-nonsense high-contrast oldstyle-serif face. It is a contemporary classic for newspaper and book setting, designed by Fred Smeijers and distributed via OurType. Available for print and Web.

Small sample of the Benton Sans typeface

Benton Sans is a Tobias Frere-Jones performance of Morris Fuller Benton’s News Gothic genre. Designed for Font Bureau, it is not only a great typeface for small print in newspapers, but one of the best-rendering text faces for the Web as well. Available for print and Web.

Small sample of the Ibis typeface

Ibis is another Font Bureau typeface, designed by Cyrus Highsmith. This square serif family is also no stranger to cross-media text-setting. Ibis works just as well whether you use it in print or on screen. Available for print and Web.

Small sample of the Ingeborg typeface

Ingeborg is modern serif family from the Viennese type and lettering powerhouse, the Typejockeys. Like any proper family should, Ingeborg has optically-sized variants for text and display settings. The display versions of the typeface can get pretty far out, too! Designer Michael Hochleitner named this typeface after his mother. Available for print and Web.

Small sample of the Ludwig typeface

Fred Smeijer’s work in contempory type design is so significant that he gets two shout-outs in my list. His Ludwig type family takes a nod from 19th century grotesques, but he does not try to sanitize their quirky forms, as so many type designers had tried to do before him. Available for print and Web.

Small sample of the Malabar typeface

This is one of the typefaces that I’ve designed. I’m somewhat partial to Malabar. Available for print and Web.

Small sample of the FF Scala Sans typeface

Martin Majoor’s FF Scala Sans has been my top go-to typeface for almost 15 years. It mixes well with the serif FF Scala type, but it’s also really great on its own. Available for print and Web.

Small sample of the URW Grotesk typeface

Of all the typefaces designed by Hermann Zapf over his long career, URW Grotesk is clearly the best. Unfortunately, it has been a little overlooked. URW Grotesk is a geometric sans, with a humanist twist that brings much more life into the letters than this genre usually allows for. Plus, the family is super big. Available for print and Web.

Small sample of the Weiß-Antiqua Typeface

Are you a DIY-fan? Do you like to print with letter press, whether you set your own type by hand, or have polymer plates made? Then check out the typefaces of Emil Rudolf Weiß! His Weiß-Antiqua is an eternal classic. Weiß may have passed away 70 years ago, but his work is still relevant. He was German, so his last name is sort of pronounced like Vice, as in Miami Vice. Available for print and Web.

Conclusion

There are many factors that play a role in typeface selection. Aside from just browsing through the available fonts that they have, or fonts that could be newly licensed for a project, designers regularly spend considerable effort determining the right typeface to complement a project’s content, or the message at hand. Understanding some of the thoughts that go into the making of text typeface—including how a typeface’s letters are fitted to each other to determine a text’s default underlying rhythm—helps lead to better informed decisions regarding what types are indeed apt, and which faces are better suited for other sorts of jobs. After having read this article, I hope you feel more comfortable with this kind of decision making, and that you will know what to look for with a font in the future.

Other Resources

For more information about choosing the right text fonts, you may be interested in the following books and Web resources:

1. Websites

2. Books

Note: A big thank you to our fabulous Typography editor, Alexander Charchar, for preparing this article.

(jvb) (il)

© Dan Reynolds for Smashing Magazine, 2012.

0
Your rating: None

  

The Web font revolution that started around two years ago has brought up a topic that many of us had merrily ignored for many years: font rendering. The newfound freedom Web fonts are giving us brings along new challenges. Choosing and using a font is not merely a stylistic issue, and it’s worth having a look at how the technology comes into play.

While we cannot change which browser and OS our website visitors use, understanding why fonts look the way they do helps us make websites that are successful and comfortable to read in every scenario. Until recently, there were only a small handful of “Web safe” fonts we could use. While offering little variety (or means of expression), these fonts were very well-crafted and specifically adjusted—or even developed—for the screen, so there was little to worry about in terms of display quality.

Now that we have a great choice of fonts that can be used on websites, it becomes clear that the translation of a design into pixels is not something that happens naturally or consistently. OS makers apply different strategies to render how typefaces are displayed, and these have evolved greatly over time (and still continue to do so). As we now look closer at fonts on screen more than ever before, we realize that the rendering of these glyphs can differ significantly between systems and font formats. What’s more, it has become clear that even well-designed fonts may not look right on Windows if they are missing one crucial added ingredient: hinting.

This article presents the mechanisms of type rendering, how they were developed, and how and why they are applied by the various operating systems and browsers—so that when it comes time to choose a font for your next project, you know what to look out for to ensure the quality of the typography is consistently high.

Rendering Strategies

Ideal shape, black-and-white and grayscale rendering
Ideal shape, black-and-white and grayscale rendering

Rasterization

In digital type, characters are designed as abstract drawings. When a text is displayed on screen, this very precise, ideal shape needs to be expressed with a more or less coarse grid of pixels. As screens turned from mere preview devices for printing output into the actual medium we read in, more and more sophisticated rendering methods were developed in order to make type on the screen easy and pleasant to read.

Black And White Rendering

The earliest method of expressing letter shapes was using black and white pixels, sometimes referred to as bi-level rendering. Printers are still based on this principle, and thanks to their high-resolution, the result is a very good representation of the design. On screen, however, the small number of available pixels does not transport the subtleties of the drawn shapes very well. And although we might not be able to see the individual pixels, the steps found in round contours are noticeable.

Grayscale Rendering

In the mid-1990′s, operating systems started employing a very smart idea. Although screens have a rather low resolution, they can control the brightness of each pixel. This allows more information to be brought into the rasterized image.

In grayscale rendering, a pixel that is on the border of the original shape becomes gray, its brightness depending on how much it is covered by the ideal black shape. As a result, the contour appears much smoother, and design details are represented. The type on screen is no longer merely about being legible—it has its own character and style.

This principle—also called antialiasing—is the same that is used when photos are resampled to a lower resolution. Our eyes and brain interpret the information contained within the gray pixels and translate it back into sharp contours, so what we perceive is fairly close to the original shape. A similar effect is at work when a relatively coarse newspaper image that can appear nicely shaded if we hold it far enough away (or similarly, in the art of Chuck Close). Recently, Gary Andrew Clarke took this to the extremes with his “Art Remixed” Series.

Subpixel Rendering

Apparently colored pixels increase the resolution
Apparently colored pixels increase the resolution.

The third generation of rendering technology is characterized by apparently colored pixels. If we take a screenshot and the edges appear red and blue when enlarged, then we know that we are looking at subpixel rendering.

On LCD screens, the red, green and blue subpixels that control the color and brightness of the pixel are located side-by-side. Since they are so small, we don’t perceive them as individual colored dots. Having a closer look at the “red” pixel marked by the white dot reveals the strategy: all subpixels are switched on or off individually, and if the rightmost subpixel of the “whitespace” happens to be a red one, then the corresponding full pixel is technically red.

Subpixel rendering on an LCD screen
Subpixel rendering on an LCD screen.

The benefits of this technique become clear if we desaturate the image. Compared to plain grayscale rendering, the resolution has tripled in horizontal direction. The position and the weight of vertical stems can be reflected even more precisely, and the text becomes crisper.

Current Implementations

For the display of text, almost all browsers rely on system rasterizers. When looking at Web font rendering, the key distinction we have to make is the operating system. However, there are differences between the browsers in terms of the support given to kerning and ligatures, as well as the application of underline position and thickness, so we cannot expect perfectly identical rendering in all browsers (even on one platform). What’s more, on Windows, the browser can have the font rendered by either of the system technologies—GDI or DirectWrite.

Before we look at these in detail, lets first get an overview of where each one is to be used:

Rendering modes used by Windows browsers
Rendering modes used by Windows browsers.

Windows

On Windows, the font format has a significant impact on the rendering. The crucial difference is between PostScript-based and TrueType-based fonts, and not the way these are brought into the browser—JavaScript vs. pure CSS, raw fonts vs. “real” Web fonts, etc. We will see identical rendering as long as the underlying font is the same.

File formats can give us a clue as to what underlying rendering technology is being used, although it’s best that one doesn’t completely rely on the naming conventions. For example, EOT and .ttf files will always contain TrueType, whereas .otf fonts are typically PostScript-based. But then there’s the wrapped format WOFF, which can contain either “flavor” of font format. So we don’t know which one it contains (and therefore, what kind of rendering may be used), just by looking at the file name. Unless you’re using EOT or .ttf files, and can be sure it’s a TrueType, more investigation when purchasing fonts is always recommended.

TrueType and PostScript fonts differ in the mathematics used to describe curves—something that rasterizers don’t care about too much—it only makes a difference for the type designer when editing the glyph shapes. What is more relevant is the different approach to hinting. PostScript fonts only contain abstract information on the location of various elements of each letter (and rely on a smart rasterizer to make sense of this), whereas TrueType fonts include very specific low-level instructions that directly control the rendering process. Curiously, however, the effective differences in rendering are not due to these differences in concept, but rather stem from Microsoft initially deciding to apply their new rendering engine only to TrueType fonts.

Windows: TrueType Fonts

TrueType font rendering with Windows grayscale
TrueType font rendering with Windows grayscale
TrueType font rendering with Windows grayscale.

On Windows XP, text is rendered as grayscale by many browsers. Although not as crisp as the subpixel rendering used by Mac OS, the letters are nicely smoothed and look great in large sizes.

TrueType font rendering with Windows GDI ClearType
TrueType font rendering with Windows GDI ClearType
TrueType font rendering with Windows GDI ClearType.

ClearType is Microsoft’s take on subpixel rendering. It was first made available for GDI, the classic Windows API. Although available in Windows XP, it is not used by all browsers. In Windows 7 and Vista, ClearType is the default, which makes it the most widely used rendering technology (if we were to consider all internet users). However, it is important to note that this applies only to TrueType-based Web fonts—GDI-ClearType is not applied to PostScript-based fonts.

One curious property of this rendering technology is that along with adopting the advantages of subpixel rendering in horizontal direction, Microsoft gave up smoothing in vertical direction entirely. So ClearType is effectively a hybrid of subpixel and black-and-white rendering. This results in steps within the contour, which is particularly noticeable in large sizes. These jaggies at the top and bottom of the curves are unpleasant, but unavoidable—even the best hinting cannot make them disappear.

For type in large sizes, ClearType is a step backwards in rendering quality. The gains in horizontal precision are not significant, while the rough contours spoil the overall result.

TrueType font rendering with DirectWrite
TrueType font rendering with DirectWrite
TrueType font rendering with DirectWrite.

The future is bright, at least in terms of Windows font rendering. In DirectWrite (the successor of GDI), Microsoft added vertical smoothing to ClearType. This new rendering mode (so far used by Internet Explorer 9), gives us smooth and precise rendering in all sizes. The main difference to Mac OS that remains is that it still tries to align contours to full pixel heights, which leads to even better rendering given that the font is well-hinted. What’s more, DirectWrite allows for subpixel positioning, which gives the characters exactly the spacing that they have been designed with, improving the overall rhythm and evenness of the texture.

Windows: PostScript Fonts

PostScript font rendering with GDI grayscale
PostScript font rendering with GDI grayscale
PostScript font rendering with GDI grayscale.

In GDI-based browsers, PostScript-based Web fonts are displayed in grayscale. Unlike the prevalent GDI-ClearType, this gives smooth contours. And unlike TrueType hints, PostScript hinting is easier to create, even automatically.

PostScript font rendering with DirectWrite
PostScript font rendering with DirectWrite
PostScript font rendering with DirectWrite.

DirectWrite not only gives smoother outlines, it also applies subpixel rendering to PostScript fonts. Unlike TrueType rendering, however, it allows for more gray pixels in order to reflect stroke weights more realistically. That makes it well-balanced, and similar to Mac OS rendering.

At some point in the future—browser makers and users will not switch as quickly as we wish—DirectWrite will succeed the older Windows rendering methods, and we will indeed be spoilt for choice between TrueType- and PostScript-based Web fonts.

Windows: Unhinted Fonts

Unhinted font rendered with grayscale
Unhinted font rendered with grayscale
Unhinted font rendered with grayscale.

In the old Windows grayscale mode, completely unhinted fonts look surprisingly good. Since the font does not “align itself” to full pixels via hinting, and the rasterizer does not enforce this either, we have a rendering that is similar to that of iOS. Unfortunately, unhinted fonts are currently not an option, as the next example shows:

Unhinted TrueType font in GDI-ClearType rendering
Unhinted TrueType font in GDI-ClearType rendering
Unhinted TrueType font in GDI-ClearType rendering.

As noted in many discussions on Web font rendering quality, GDI-ClearType is extremely dependent on good hinting. Horizontal strokes have to be precisely defined by means of hinting, otherwise they might be rendered in an inappropriate thickness. Even in larger sizes, hinting is crucial. Unhinted fonts will show “warts” sticking out where contours are not correctly aligned to the pixel grid, like in the example above.

Unhinted font rendered with DirectWrite
Unhinted font rendered with DirectWrite
Unhinted font rendered with DirectWrite.

In DirectWrite, unhinted PostScript and TrueType-based Web fonts show virtually the same rendering. Text fonts of either flavor will still need good hinting in order to keep the strokes crisp and consistent. Display fonts may even get away with sloppy or no hinting, since this does not show much in large sizes.

Mac OS X

Font rendering in Mac OS X
Font rendering in Mac OS X
Font rendering in Mac OS X.

On Mac OS, all browsers use the Quartz rendering engine. TrueType and PostScript fonts are rendered in exactly the same way, since hinting—the biggest conceptual difference between the two formats—is ignored. The subpixel rendering on Mac OS is very robust, so this platform is typically the one we need to worry about the least. The rasterizer doesn’t try to understand the strokes and features that make up a font, as everything is represented by more or less dark pixels. Since the letter shapes are not interpreted, they cannot be misinterpreted. Quartz rendering is reliable because it doesn’t try to be smart. As a side note, Apple does seem to apply some subtle automagic to enhance the rendering, but this is entirely undocumented and beyond our control.

In some cases, however, this leads to less-than-ideal results. In the above example, the large size “T” has a fuzzy gray row of pixels on top because the theoretical height is not a full pixel value, and Mac OS does not force its alignment. Unfortunately this cannot be controlled by the font maker. However, the blurriness occurs only in certain type sizes. So typically, choosing a slightly different font size fixes the problem. With a bit of trial-and-error, one can find a type size that looks comfortable and crisp.

Another difficult-to-control phenomenon is that on the Mac, type tends to be rendered too heavy. This difference is most noticeable in text sizes, where the same font can look a bit “sticky” on Mac OS while appearing almost underweight on Windows.

iOS

Font rendering in iOS
Font rendering in iOS
Font rendering in iOS.

The rendering on iOS follows the same principles as on Mac OS—the main difference is that it currently does not employ subpixel rendering. The reason might be that when the device is rotated, the system would have to re-think and update the rendering because the subpixels are physically oriented in a different way, and the makers wanted to minimize CPU use.

Conclusions

Website visitors use a great variety of systems and browsers. Some are not up-to-date, and sometimes it’s not even the user’s fault, but rather a company’s policy to stick with a certain setup. My personal opinion is that we should try and give them the best rendering we can, instead of blaming OS makers, or demanding users to switch to better systems.

On Mac OS and iOS, we hardly have any control over the rendering, which is acceptable (since it’s generally very reliable). One problem is that fonts generally render too heavy. Maybe some day, Web font services can improve the consistency by serving slightly heavier or lighter fonts depending on the platform.

On Windows, hinting matters—especially for TrueType-based fonts (the only Web fonts Internet Explorer 6–8 will accept). Apart from that, one significant control we have over the rendering is the choice between TrueType and PostScript. Except for very well-hinted fonts in smaller sizes, the latter is equal or superior in rendering, and easier to produce. Even though DirectWrite is making Windows rendering more pleasant, it will not remove the necessity to provide well-hinted fonts.

Practical Application: Improving Display Font Rendering

Some Web font providers (such as Typekit or Just Another Foundry), have started serving display fonts in PostScript-based formats.

JAF Domus Titling Web rendered with GDI ClearType
JAF Domus Titling Web rendered with DirectWrite
JAF Domus Titling Web rendered with Windows grayscale
JAF Domus Titling Web in Mac OS X
JAF Domus Titling in different rendering environments.

While the GDI ClearType jaggies are unavoidable for IE 6–8, all other scenarios produce nice, smooth results. This also means that we will still need fonts that have decent TT-hinting—the browser share of IE6–8 is still too big to deliver fonts that don’t at least render in a clean fashion.

Bello by Underware on Typekit
Bello—by Underware on Typekit—served as PostScript-based Web fonts (right), which gives smoother rendering than TrueType (left).

Typekit has also started to implement a hybrid strategy by serving display fonts as PostScript in order to trigger smoother rendering in Windows GDI. This requires some decisions to be made on the basis of visual judgement.

“How do you define a display font?”, you may ask, and it is indeed difficult to draw the line. Some of the foundries offer high-quality, manually hinted TrueType fonts that look great in text sizes (and it would be a pity to lose this sophistication by converting them to PostScript). Some text fonts may well be used in very large sizes. So ideally, we would have to offer them in two different formats. However, increased complexity of the UI (as well as back-end handling) have so far kept us from doing this.

Future Developments

More and more type designers are becoming aware of the technical issues that arise when fonts are used on the Web, particularly TrueType hinting. As the Web font business grows, they are willing to put some effort into screen-optimizing their fonts. In the near future, we will hopefully see a number of well-crafted new releases (or at least updates to existing fonts).

With increasing display resolutions— and more importantly, improving rasterizers—we will slowly have to worry less about the technical aspects of font rendering. GDI-based browsers will certainly be the boat anchor in this respect, so we won’t be able to use TrueType fonts that aren’t carefully hinted for yet another few years. Once this portion of Web users has become small enough, the process of TrueType hinting (which is time-consuming and requires considerable technical skills), becomes less crucial. While most Web fonts currently on the market are TrueType-flavored, I am expecting that the industry will largely switch to PostScript, which is the native format nearly all type designers work in (the fonts that are easier to produce).

Other Resources

(jvb) (ac) (il)

Note: A big thank you to our fabulous Typography editor, Alexander Charchar, for preparing this article.

© Tim Ahrens for Smashing Magazine, 2012.

0
Your rating: None