Design critique of Tiresias Screenfont

Elsewhere, we pretty much blew out of the water all of the so-called “research” that claims to justify the use of Tiresias Screenfont for captioning. But how successful is the design of the font? (See full character showing [PDF].)

“Amateur” in one sense

In no way could Tiresias be considered an amateur design save for one: Apparently no one got paid to design it. John Gill separately told us that the design team had “15 people in it, all of them unpaid.” The closest thing to a type designer anywhere in the gladhanding documentation of this near-vapourware project are Chris Sharville (of Laker-Sharville) and Peter O’Donnell. We assume that qualified type designers worked for nothing on this font because the ostensible beneficiaries were disabled and the commissioner was a charity.

We don’t see this as a viable model for developing fonts for captions and subtitles; the reading demands of those media are too precise and exacting to be borne on the backs of designers’ free time. People with disabilities are too important to rely on charity. Besides, if we’re paying you we can demand better work.

Nevertheless, even working for nothing it is unlikely that experienced type designers could colossally screw up, and indeed that did not happen with Tiresias. On first blush it seems to be a viable font, and many features obviously have been well considered. The problem is that the designers are not being graded on intentions or general success; the criterion of interest is actual suitability to caption and subtitle use, and that means all the details count, not just the ones they got right. And “actual suitability” applies to all caption and subtitle viewers, not just the low-vision grannies who dominated the testing phase of this typeface’s development.

General features

Tiresias sets loose at high resolution but sets too tight in bitmap usage. When you’re reading Tiresias captions you’re reading bitmaps, and the default spacing is too close. Admittedly this is difficult to finesse in media like DVD and DVB that do not have subpixel rendering, meaning that letterspacing is counted in units of full pixels. But the fact remains that captions are luminous, thus they glow and blend into each other. Letters spaced one or two pixels apart begin to smudge.
There is no italic. This deficiency, unimaginable in a 20th-century text typeface, was inexplicably defended by John Gill, who stated that “USA subtitling requires italics and bold which are not used in European subtitling.” Italics aren’t used because the old teletext fonts didn’t have them. It’s like saying italics aren’t needed because manual typewriters can’t produce them. Italics are necessary for rendering the English language, not to mention other languages. But, in a classic example of a skiamorph, a previous limitation was perpetuated in a new medium that never needed to have it. The simple lack of italic is enough in itself to reject Tiresias from consideration.

Confusable shapes

Some clear effort went into differentiating oft-confused shapes:

I has no crossbars, l has a curled tail, 1 has no base and an angled bar. (Vertical bar | is as hard to differentiate from one or more of those other three characters as it ever is, though admittedly its use is rare.)
Brackets [ ] have descenders but are narrowly sized for print, not display. These characters, along with parentheses, simply cannot be too subtle for captioning use. Remember, some captioning shops set full spaces inside of brackets –
- [ "I ALONE" PLAYS ON JUKEBOX ]
- [ RING ]
– while others run them solid with all-upper-case text. The result is confusion with the English word I.
Classic confusable numerals have different shapes, but arguably not different enough. 5 and 6 are more distinguishable than usual in sansserif fonts, but 6 and 9 are mirror images and 8 has many of the features of 6. B doesn’t look at all like 8 given that B’s upper and lower bowls have different widths. (This reduces the Bness of the letter, arguably beyond the point of acceptability.)
Further mirror images are found with M and W, an inexplicable choice given that M can have vertical sides and W cannot. In their canonical forms these letters are distinguishable; Tiresias turns them into a dyslexic’s nightmare. They’re also too wide in a face with a generally condensed aspect.
a, e, and o are differentiable, but the distinguishing features pretty much boil down to the unusually narrow top half of the bicameral a.
K is symmetrical around the midline.
Single and double quotation marks are angled slabs in 6/9 orientation, hence are confusable (and even confusable with the neutral forms). If you have to use angled slabs, use mirror-image-of-9/9 orientation, like Verdana.

Unnatural forms

Tiresias falls down completely on a few notable letters.

The typeface has the worst-rendered J and j in living memory. J descends too far and then has a hook. It looks dreadful at the beginning of a word and even worse at the beginning of a line (particularly in a left-justified flush-left caption). The hook on the j extends too far left and seriously clashes with y. Esperanto J plus circumflex (Ĵ) borders on a joke. On the plus side, Ĳ ĳ both kern nicely.
Crossbar of the G extends just a hair too far to the left, resulting in misrendering in DVD bitmaps. Maybe it could have used a Helvetica- or News Gothic–style spur.
It’s not clear how a B can be wider than a G in any professionally-developed font, but that’s what we have here.
f is the right width but t is too wide, resulting in ungainly appearance for words like “gift,” “Hatfield,” or “ofttimes” (or many German words, like Luftfeuchtigkeit ‘humidity’).

Accented forms

The narrow top half of the a looks unpleasant with dieresis and tilde (ä ã), which extend past the left edge of the top stroke. Left and right halves of æ are asymmetrical, but this may actually assist in character recognition. a plus ogonek (ą) extends too far to the right (and ogonek stroke thickness seems too great).
Since the i is just a straight line with a dot (and no left-extending crossbar at the top of the stroke), we have the classic problem of wide accents on a narrow letter (ï î) and the related problem of collisions and near-collisions with adjoining letters, which even happens with acute and grave (í ì). Another deficiency of this form of i is dotless i (ı), which designers tend to overlook and for which we here at Screenfont have no obviously superior alternative.
t plus bar (ŧ) looks like a mistake or a pound-sterling sign.

Pi characters

¢ $ are too bold.
Check mark ✓ and ballot X ✗ appear to have been taken directly from Zapf Dingbats or Wingdings and don’t match the rest of the font.
A much-used character in NTSC and ATSC captioning, the staffnote ♪, appears to be absent, but the beamed version ♫ is there. This may represent a British misunderstanding of U.S. and Canadian captioning, though it probably represents a British attempt to imperiously and unilaterally fix or “improve” U.S. and Canadian captioning (along with renaming it “subtitling”).

Can Tiresias be salvaged?

No. And in fact things are getting worse, since sources tell us that a Cyrillic version has been commissioned, again without italics.

Version history

2006.02.16 16:24: Posted.

You were here: Screenfont.ca → Fonts → Today’s fonts →
Tiresias → Design critique of Tiresias Screenfont