What fonts should we use in the meantime?

So: If the fonts we’re presently using for captioning and subtitling are complete shite – especially Arial – what should we be using instead?

And if those interim fonts are good enough to replace the existing crapola fonts – especially Arial – then why should we wait around for this project to design, test, and sell or distribute its own sets of fonts?

On this page

  1. What’s wrong with Arial?
  2. What’s worse than Arial?
  3. Spacing and pixels
  4. What’s better? Some other fonts you could consider
  5. Take our word for it
  6. Why wait for even better fonts if these fonts are so much better than Arial?

We’ve also got illustrations on a separate page and a discussion of the history of grotesk faces like Arial and Helvetica.

What’s wrong with Arial?

  1. Without question the most important reason not to use Arial is because it makes you look like a rube.
    1. It’s closely associated with Microsoft Windows, and with the bad type that Windows all but irresistibly produces.
    2. Arial is a two-bit knockoff of Helvetica, with a range of design inconsistencies.
    3. Those inconsistencies make it the black sheep of a grand and venerable family, the grotesks (spell it with a -que if you like). Forrest Novell defines grotesks as having “rational, legible forms; low contrast; consistent stroke widths (although [not] generally monoline); a strong vertical emphasis.”
  2. You already have a range of well-designed, internally-consistent, historically-significant grotesks at your disposal if you demonstrably need them, including Helvetica (history), Univers (history), and perhaps most pleasing of all, Akzidenz.
  3. But grotesks are unsuited to captioning and subtitling because of the many confusable character shapes. In fact, all the classic confusable combinations – Il1|, S568, rn m, cl d – are genuinely confusable in grotesks.
  4. Classic reversible characters (the bdqp-and-sometimes-g set, which you can flip and rotate to form other letters) are indeed all reversible, making grotesks an unnecessary challenge for dyslexics and people with other learning disabilities.
  5. The geometric letterforms – in which letters like Cc, Oo, Q, and e are constructed from near-perfect circles – generate their own confusions. But, more importantly, they perform poorly at the low resolutions typically found in captions and subtitles. (Indeed, almost every caption and subtitle medium has a low resolution, other than the rare and costly example of type burned into film prints via an optical process.)
  6. Some of Arial’s design features get in its own way in caption/subtitle use. The weirdly extending, weirdly wavy bar of the 1 and the unnecessary point of the stem of the t render poorly at low resolution, making the 1 too wide and lopsided and the t distractingly pointy. (Have a look at it yourself. Since the point at the top of that t is one of the differentiators between Arial and Helvetica, hinted bitmaps and other systems go to great lengths to preserve it. You’ve always got a stray pixel or two at the top of what should be a leveled-off stem. The pointed t is, quite simply, an error.)
  7. A perennial issue in sansserif fonts whose Ii and l are narrow is what happens when you add diacritics or accents, as in these examples from French and Polish.
    • aïeux
    • exiguïté
    • haït
    • bélître
    • paraîtrait
    • dîner
    • Île
    • discoïde
    • abecadło
    • bałtycki
    • błysk
    • fiołki
    • półfinał
    • półhektar
    • półlitrowy
    • zżółkłbym
    • żółw
    This isn’t Arial’s fault specifically, but you avoid or lessen the problem if you don’t use Arial. (See Gaultney’s paper “Problems of diacritic design for Latin-script text faces.”)

What’s worse than Arial?

Of course you can do worse than use Arial. Real-world examples? Subtitles that appear to have been typewritten in Courier and then imposed on the film stock. MusiquePlus’s bizarro monospaced Futura knockoff is a stunning example of just how tacky Quebeckers can be; it’s more Sexe-Si-Bon than Outremont.

But really, the typeface in common use for captions and subtitles that is notably worse than Arial is Arial Narrow, a bastardized pseudocondensed variant that was just as repellent in its original form, Helvetica Narrow.

First of all, why do you need a condensed font for your captions? Trying to jam a few more words in there? That’s fine on the surface, but there are natural readability limits, and if you need a condensed font you should use a real condensed font, not a regular font whose letters were scrunched. (Like a fat person trying on a too-tight pair of pants, curves end up squishing outward in the most inopportune places.)

If you use Arial you might just be excusably ignorant, but if you use Arial Narrow you’re lower than dirt.

Spacing and pixels

A typical problem in film titling must be emphasized here: The default letterspacing is too tight for video. It may work just fine in print, even in laser printing, but we simply have too few pixels to work with in video. Letters are too close together and, because letters are often luminous, halation causes letters to run together even more.

It’s quite indicative of the level of disinterest in typography found among developers of captioning and subtitling software that spacing is so difficult to control, in most cases impossible. But you, the software developer, can read it, which means there’s no problem, right? No, there are lots of problems. It’s just that you don’t know what the problems are, so you don’t give the user a set of controls to deal with those problems.

There’s another detail: The DVD format’s rectangular pixels. (They are errantly and infuriatingly called “non-square pixels”; what are they, octagons?) Rectangular pixels make it rather difficult to draw a nice clean circle, as found in geometric sansserifs and a lot of fonts that contain marginal, seldom-used characters like o and e.

If your computer system assumes square pixels, as Windows and Macintosh systems do, then your letters will appear scrunched and too tall when the same number of square pixels are transformed into upright rectangles. Essentially, every font you use becomes faux-condensed.

Are we done yet? Nope. Rectangular pixels also make italics hard to render properly. There’s a reason italics in DVD subpictures look so jagged, and it’s not just because we don’t have enough pixels and anti-aliasing is all but impossible.

What’s better?

Stop using Arial right now. For most caption and subtitle applications, don’t use any of the other grotesks, either. That means no Helvetica. (In fact, it’s hard to imagine a caption or subtitle application where grotesks should be used.)

Go out and buy at least the roman and italic of one or more of these fonts:


It’s somewhat overused and is now a bit of an old chestnut, but the advantages built into the font’s intended usage – as implied by the name, Officina was designed for office documents produced on medium-resolution printers – are advantages that work well in captioning and subtitling. Don’t buy the version with ranging (old-style) figures.

Few confusable character shapes; dramatically legible numerals; has quite a bit of personality without being oddball (consider the g, f, R, S); sets narrow; small caps available.
Surprisingly bold even in the book weights; like most fonts, letterspacing is too tight for captions and subtitles; even the degree of personality it has can be too much for some material, like dramatic productions set in the past; too closely resembles typewriting even though it’s proportionately-spaced; annoying name.

You’d probably want Officina Sans Book, by the way, or perhaps Medium.

The Serif variants would probably not work as well, but they’re an interesting case in that they are essentially slabserifs, which we’ll cover below. You could give them a try, but you’ll need even more letterspacing than usual.


Not the most exciting sansserif face in creation, but then, it’s not meant to be. Unit is a “stricter” variant of Meta, we are told, which may be another way of saying “Meta without idiosyncracies.” (Meta, by the way, is unsuitable for captioning and subtitling due to the unjoined bowl in the lower-case g.)

Doesn’t have Meta’s unjoined bowl; curved foot of l, but straight-l variant available; generic in possibly the right way
Light is too light, medium too bold; narrow f and tall t – two problems we’re trying to avoid by not using grotesks in the first place; confusable O, Q


The font with the very worst name in the world. More variations available than Unit, though alternate character forms (like l with or without tail) are not available.

Slightly fewer confusable characters; discernible numerals
Not much better than Unit, really; numerals are discernible because they are somewhat oddly shaped

Certainly avoid the serif and slab versions, whose haphazard serifs simply would not work in blocks of text that viewers are expected to read instantaneously.


You’d need a specialized application for Luc(as) de Groot’s Taz and Taz Text. The families are related, with Taz looking more like a Futura Condensed (with its single-bowl a) and Taz Text looking more like a variant of Meta.

Legible; vast character set, as ever with Luc(as); unconfusable O, Q
Taz Text has only ranging or old-style figures. Angled stroke terminals resolve poorly on coarse screens. y and g are fussy and distracting, with their angled, unterminated tails


So far, the recommended typefaces have all been sansserifs. Untrained commentators tell us that serif typefaces are easier to read (generally but not always true), or they tell us that serif typefaces work badly on screens (ditto). But in both cases, these untrained commentators are assuming there’s only one kind of serif typeface – something with delicate little blobs at the end of strokes, like the only example they can ever think of, Times. (Or Times New Roman, as they often call it, revealing a telltale Windows bias.)

Some traditional serif fonts look just fine on LCD screens powered by operating systems that use subpixel rendering, like Windows ClearType and Mac OS X. They look pretty crappy on computer screens in other contexts, at least at text sizes. Those facts are noncontroversial.

But a subset of serif fonts, the slabserifs, work shockingly well in low-resolution environments, including computer screens and captions and subtitles. Slabserifs are so named because the serifs are unusually long and sit at right angles (or nearly so) to the underlying stroke. The thickness of the serifs is similar to that of the stem it projects from.

If you used a ballpoint pen and weren’t allowed to retrace your strokes, the kind of serif font you would draw would probably be a slabserif. They’re seemingly unsubtle, but many such faces are surprisingly legible. And the tidy right angles translate easily to low-resolution devices like computer screens.

Moreover, the famous custom “subtitling” font designed by the Department of Typography at the University of Reading, now in use for over 20 years, is an actual slabserif face and it works remarkably well.

You might want to disregard geometric slabserifs like Lubalin Graph (though the Condensed variant, designed decades later with Avant Garde Gothic Condensed as a base, could work). But a few other slabserifs to consider are Serifa, Rockwell, Lucida Fax, and possibly TheSerif, if only in the Condensed variant. (TheSerif is one of the over 300 variants of Luc[as] de Groot’s Thesis family. Its spiritual ancestor, Cæcilia, is possibly too fussy for caption and subtitle use.)


You may be wondering if this is rather too much detail for something as evanscent as typefaces. Your response to the whole question might be “I can read it and nobody’s complained yet.” While common, that response is sadly irrelevant.

We’re not talking about your personal, usually untrained, responses to a certain typeface; you’re creating captions or subtitles for your audience, who have needs you may not have. Complaining is nearly impossible in the first place (who do you complain to at 10:30 on a Sunday night while you’re watching a video?); bad type may lead to misreadings and discomfort, but it’s so subtle that only experts have the vocabulary to talk about it in the first place.

You expect your clients to accept that they don’t have the expertise to caption or subtitle their own productions. They need you for that, because you’re the experts. Well, we’re the experts on type. Take our word for it.

Why wait for even better fonts?

All right. If’s métier is the design, testing, and sale and distribution of improved fonts for captions and subtitles, why are we giving you a whole raft of substitute fonts you can use in the meantime while we get our act together?

  1. You urgently need something less bad than what you are presently using.
  2. Tiresias isn’t any good, either, but we’ll save that for another day. (Hint: The design isn’t very good, there’s no italic, it wasn’t tested save for a tiny group of visually-impaired people, and it costs up to $12,000.)
  3. There aren’t any fonts custom-designed for captioning and subtitling and tested to prove they work. Even the fonts we’re suggesting here are a kind of kluge.
  4. All these interim fonts have a few pros in their favour, but also significant cons. We aim to produce fonts with few cons or none whatsoever – and back it up with studies.
  5. If everyone took our advice about interim fonts, everyone’s captions and subtitles would end up looking the same. We, on the other hand, are in favour of custom and exclusive fonts for certain applications and clients.

But at root, we are asking you to take us at our word that our candidate fonts will be visibly better than anything we’ve got now. While we’re waiting, put the Typocratic oath into effect and first do no harm. Try some of the interim fonts we’ve suggested. Try something – just stop using Arial.

See for yourself

You can have a look at some illustrations of Arial in use on a separate page.

Version history

Added a discussion of the history of grotesk faces like Arial and Helvetica.
Moved illustrations to their own page.

You were here: Screenfont.caFontsToday’s fonts
What fonts should we use in the meantime?