Know any good phonetic symbols fonts?

axeltomas's picture

We're in the preparation phase of a project where we will be designing an extensive grammar book. As expected there will be a lot of phonetic transcriptions. Book design is comfortable territory but we have never done anything using a font with phonetic symbols.

Anybody got experience with phonetic fonts? Please let us know of any good, complete font families with phonetic symbols.

Jongseong's picture

Check out International Phonetic Alphabet and the following discussions in particular:

Serif Fonts with IPA Characters for Book Work
Phonetic fonts

Will you be setting the phonetic symbols in the same typeface as the rest of the text? What sort of typeface do you plan to use for the rest of the text? That will help narrow things down.

axeltomas's picture

Thanks Jongseong for the quick and helpful post.

As said, we are in a very premature phase so nothing is decided but trying to answer your questions I'd say that the Phonetic transcriptions don't necessarily have to be in the same typeface as the rest of the book. As for the rest of the text we are considering the use of MERLO by Mario Feliciano or some other good book face.

Jongseong's picture

A sans serif for phonetic transcription might be a good idea. I can't immediately think of anything that goes really well with Merlo, though. But look through the lists to see if you can find something. Also note that some fonts that come with Windows Vista like Microsoft Sans Serif and Tahoma have been extended to include phonetic symbols.

Alternatively, it might be worth contacting Mário Feliciano to see if he will be open to adding phonetic symbols to Merlo.

charles ellertson's picture

Since one of the threads listed was one I started, I might as well finish the tale. I could find no suitable commercial font that had all the needed characters, except, as noted, Times and Stone.

Since Adobe allows modifications to their fonts by the end user for their own use, I used Minion, primarily because of its neutral feel, and because it does come with Greek (though you cannot simply use the existing Greek characters for all the IPA "Greek").

I then made up the needed phonetic characters, roman and italic.

While IPA may not require italic, try telling that to an editor who has been taught that all foriegn words must be set in italic. Many editorial decisions are religious in character, that is, they are reveled truths, brooking no argument. Italic is needed.

It does not take all that long to draw up serviceable IPA symbols -- maybe a week for italic and roman. Note the word "serviceable." They probably wouldn't pass muster with a real font designer. But at text size, they worked well enough.

* * *

I have used Merlo in one book. It tends towards being too heavy, and I'm one of the guys who complains that most photocomp and digital fonts are too light. With a 10-point setting, it worked pretty well. With a larger setting, I doubt it -- a noted designer (Rich Hendel) tried a larger setting size and couldn't get it to work. That said, I will use it again, when it fits the project.

* * *

In conclusion, I would use a font from a foundry that allows the end user to modify the font for their own work, and make up your own symbols, as needed.

Jongseong's picture

That's great advice from Charles. Depending on the scope and phonetic detail explored in the grammar book, you will probably need only a small number of additional phonetic symbols (around a dozen for Portuguese), not the entire IPA. Having the needed glyphs custom-made will not take too much time.

Also, many dictionaries, including the Longman Pronunciation Dictionary, use the convention of setting the symbols for sounds that may be optionally dropped in italic. So you can never say IPA italics are not necessary.

guifa's picture

Also it's not unheard of to have a section title or that has a minimal amount of IPA,, and if the style guideline has that placed in italic, then logically the IPA would also go in italic. *secretly happy that his font's italic a was always designed to be two story making IPA italic much easier*

charles ellertson's picture

Yea, drawing up (stealing parts) to make a two-story italic "a" is one of the tough ones. For the one-story Roman a, think "c" plus "iota" with a little adjustment, remove overlap, sidebearing adjustments, and on to the next.

.00's picture

Another option would be to identify the commercial fonts you are interested in and see if the designer is interested in expanding the character set for your needs. It may cost you a few bucks, but the results would most likely be better than hammering away on your own.

raph's picture

You might take a look at Junicode also. It's in the same general family as Merlo, but I have a feeling the pair won't actually work well in practice (Merlo for text, Junicode for phonetics), because the former seems much heavier than the latter.

For a Times-like font with IPA characters, take a look at Doulos SIL.

Both of these are free, and so actively encourage you to modify them for your needs. You might be able to make slight weight adjustments, for example, in your favorite font editor using a script.

Best of luck!

Andreas Stötzner's picture

The entire IPA and UPA repertoire (blocks 0250, 1D00, 1D80; all phon. char.s)
*complete* in Regular *and* Italics:

charles ellertson's picture

the results would most likely be better than hammering away on your own

James, I take mild offense at that remark. And the "few bucks" always seems to really be a "few *hundred* bucks" ("few" being "more than three").

@Ralph, I'd probably try Charter from SIL before Junicode. Junicode is far too light, and a bit too expanded for Merlo. In their current form, and for different reasons, both Charter and Junicode are what I would consider good fonts for manuscripts, but not books,

guifa's picture

Andreas, the italic version of that font has some important problems for use in IPA, most especially it lacks a two-story a, and its turned version. And it's missing a good number of suprasegmentals.

.00's picture

Oh, get over it Charles! Of course it is a few hundred bucks. How much is your time worth? So if you do it your self it has no value, but if the type designer does it for you it becomes a rip off?

I always go back to what my friends in the Peace Corp used to tell me, there are only three numbers in the world: one, two, many!

And just to snap your head around, we recently responded to a client request and did a Cyrillic upgrade for two of our fonts FOR FREE!. Simple because we were interested in the project.

Take your mild offense and ...

gaultney's picture

Sorry for the self-promotion, but no one has yet mentioned Charis SIL - full set of styles and complete phonetic coverage with a design similar to Matthew Carter's Charter:

Gentium Basic and Gentium Book Basic also are complete families, although their character set support is only 'basic'. They still support a good range of IPA and phonetic characters, which may be sufficient for your needs. The Book weights, in particular, are worth a try, and support smart diacritic positioning via OpenType:

Andreas Stötzner's picture

… it lacks a two-story a, and its turned version. And it’s missing a good number of suprasegmentals.

Sorry, I don’t quite understand. This is the first time I hear such complaint. Could you be a bit more descriptive?

guifa's picture

In IPA, there are four different versions of the letter A (two forms and their respective rotated version). As you know, it's standard in serif fonts for the lowercase a to change form between normal and italic faces. This is a contrastive element in the IPA alphabet, so it requires this italic single-story a to be present in the regular upright face. Just the same, in an italic font, the upright two-story a most be designed in italic as such and cannot revert to a single story form.

Otherwise, a word like father in English when transcribed will become ambiguous:

faðɜɾ = Scottish
fɑðɛɾ = Irish

In Andron mega, if I set this in Italic, BOTH the Scottish and Irish transcription will appear identical which shouldn't happen. The suprasegmentals are those symbols that indicate tonality, stress, and length and are in the spacing modifier letters block. These are an integral part of IPA. The English word for pot (SAE) should be pʰɑt, which uses the raised h. Any tonal language needs some of the other ones, and some people might need occasionally the double tonal marks found in the 1DC0 block.

charles ellertson's picture

Take your mild offense and ...

What I find hard to take isn't so much the cost, though there are times when and extra $1,000 (20 glyphs, each in roman & italic, at $25 per) will kill a project.

Rather it is the notion that the type designer is the ultimate authority, as opposed to the several end users. About once a month, we get someone, an author, an editor, or a designer, who wants something typographic changed. There have been so many of these instances over the years I can't keep track.

This week, for example, proof came back where the publisher (probably the author, but who knows), wanted the breath marks on Greek capitals placed over the letter, rather than in front. I could tell them they are wrong. Or, I could admit that Greek began around 1,000 BC, and has undergone many changes in how it is written. Maybe at some point in time, that's how they were written. I'm certainly no expert on the permutations of the Greek language. Or we could just set what they wanted, which is what we did.

There were only a few occurrences, so we made the change using the layout program. If there had been a lot of instances, I would have gone into the font, created a separate variant, and written a feature to switch them on. The type designer would not have been consulted. Did I mention that second proof is usually due 10 working days from when it is received?

I have also had authors insist that certain phonetic characters appear certain ways, in spite of the "guidance" about allowable variance that can be found in Chicago's Phonetic Symbol Guide. Again, you usually find this out when first proof is returned.

Type isn't hung on a museum wall, and OpenType software can't cover all the users needs.

If you want to put a clause in your EULA that your fonts cannot be modified, that's your right. It does affect their usefulness.

abi's picture

There is a phonetic companion to Fedra Serif.

axeltomas's picture

I've been caught up in meetings and work with little or no time free. Just a short burst to thank everybody for the great contributions and I will post back with developments asap.

Andreas Stötzner's picture

“four different versions of the letter A”

Of course I know this problem. It is actually unsolvable within a family which contains phonetic Italics. As you can see I applied the trick to make the 0251 and 0252 glyphs in the shape of the ‘fishy’ alpha in Italics to avoid this particular ambiguity. Someone once suggested this, don’t know anymore who it was. Yes, the one-storey a at 0250 in Italics corresponding to the two-storey a in Reg. is a mess for phonetics. But what you simply *can’t* do is therefore change the italic a at 0061 into a two-storey glyph. A two-storey italic a is rubbish anyway. But “scientists” like to invent such typographical nonsense and then leave us with the problem “BUT I WANT IT NOT THIS WAY, BÄÄÄH, I WANT IT OTHERWAYROUND MÄÄ MÄÄ MÄÄ!!“ Really anoying.
Seriously: HOW to deal with that sensibly? Any known best practice, recommendations? NO, no Opentype excuses, please.

“In Andron mega, if I set this in Italic”

Well, you can do *very* much in Andron Mega, but perhaps not everything. To say the least, it’s not the typeface being wrong here, it’s the script system itself bearing the fault.

The suprasegmentals …”

I’m not sure wether we are talking about the same thing, but all 1DC0-characters are there in Andron Regular.

However, thanks for your remarks. If any convincing solution shall be feasible for the a-problem I’ll be happy to take care of it with the next upgrade of Andron.

Jongseong's picture

But what you simply *can’t* do is therefore change the italic a at 0061 into a two-storey glyph.

Why not? I am not saying that the italic a for general uses should be a two-storey glyph. I am saying that for use in phonetic symbols only, you could provide the two-storey variant for the italic a. That seems to me the most reasonable solution. Just because one would generally object to two-storey italic as we shouldn't object to such a solution for a specialized use.

You can't limit yourself to one glyph for one codepoint and expect it to work for all situations.

By the way, I know that at least one phonetician, David Wells, has objected to the single-storey a as a phonetic symbol looking like a 'fish-shaped alpha', although he was talking about the roman, not the italic. I don't think that line of disambiguation in the italic works for phonetic purposes.

charles ellertson's picture

Brian, the problem with having a two-story "a" in italic is there is no code point for it -- outside, of course, the regular "a" in 0061, which is usually a one-story a.

So what to do? Switch it with an APPH or IPPH language tag feature? But not all a's in an italic sentence should be switched. More to the point, I have never seen a manuscript with any language tags, let alone an IPPH tag. Authors and editors just don't use them.

We are left with having it as an unencoded glyph, to be found by the unsuspecting comp. Or a PUA glyph, equally hidden.

Jongseong's picture

Charles, that is true. The two-story italic a would be an un-encoded alternate glyph. Theoretically, one could set it up so that it would be used only for text tagged as IPA. However, support for this feature is generally absent and as you say, no one uses the language tags anyway.

It will be up to the editor to make the substitution manually. I cannot imagine that there ever would be a text using the two-story italic a extensively; I've never seen an example myself, and I've seen a lot of linguistics texts.

However, the Fe'fe' or Nufi language of Western Cameroon uses both a and ɑ in its alphabet, so setting a text in that language may require more substitutions than an editor could handle. In that case, assuming the entire text is in that language, modifying the font so that the italic a is the two-storey version would be the simplest way.

John Hudson says he recommends having a dedicated IPA font, since there are examples where IPA symbols share code points with letters used more often in other contexts (like the Greek letters) that require design differences according to use. Distinguishing the italic as in IPA would be another example to support such an argument.

charles ellertson's picture

A separate IPA font is one way to go. Unfortunately, this means there is no kerning with the non-IPA font -- unless you have all the normal glyphs in the IPA font, which sort of defeats the purpose of "separate," unless all you were worried about were non-Latin characters.

The problem is exacerbated by the fact that most romans use a two-story *a*, most italics a single story *a*, which means InDesign's "character style" probably won't work as a solution. It is just going to be handwork. BTW, I have seen several fonts where the nominal "a" in italic was two-storied, I just can't think of any off-hand. And that's not a solution, anyway.

Jongseong's picture

I am talking about having a separate IPA font that shares the exact same glyphs with the regular font for the usual Latin characters, while having the Greek letters for example optimized stylistically for IPA, not regular Greek text.

Yes, the font substitution is all going to be handwork. Luckily, you would never have to kern between the IPA font and the non-IPA font, since a phonetic transcription would be separated from the surrounding regular text by brackets or space.

Peter Farago's picture

Hi, everyone. I'm a graduate student in linguistics and a bit of a type geek, and I've registered an account here to throw in my two cents this issue. In general, I very much agree with Jongseong. The differences between the various "a" characters in IPA is meaningful, and probably the only way to effectively handle it in italics is with either a separate font, or an OpenType or AAT Stylistic Variant (sorry, Andreas). As an end-user, I would personally favor the second approach, although I have no idea how much work is entailed in implementing such a feature.

Charles_e's comment about fussy editors demanding italicized IPA is dead on, as I've experienced such things myself. In addition, there are a great many minority languages which have developed orthographies which include phonetic characters. When it becomes necessary to cite words or phrases from these languages, italic styles are practically indispensable. Unfortunately, there are very, very few acceptable IPA faces with italics, and none at all which include a matching typeface suitable for body text. This is one of the reasons I find Andron so exciting: it's an actually nice-looking typeface that features a nearly complete set of IPA characters. The requisite two-story variants of 0061 (a), 0252 (ɐ), and 00E6 (æ) (which, although not discussed above, becomes hard to distinguish from 0153 (œ) in most italic faces) would make it perfect.

Andreas, I share your frustration that the designers of the IPA seemingly failed to consider some of the realities of type design, but the truth is that we live in an imperfect world, and we often have to live with the poor decisions made by our predecessors. It adds some complexity and challenge to life, and I think there are more than a few linguists out there who would appreciate your taking the trouble to work out typographic oddities like the italic two story "a".

Jongseong's picture

Thank you, Peter, for the valuable end-user input. When you say you would prefer the stylistic variant approach, is that because you already work with OpenType- or AAT-savvy software? Or do you just mean that software improvement also on your wish-list, along with the IPA fonts designed to take full advantage of them?

In general, would linguists and other end-users of IPA fonts have the software to handle the stylistic variants? In your experience, how would they handle these types of issues? My guess is that lots of them do not have the capable software, and that's why separate IPA fonts would be a reasonable solution for now. But I would agree that having the IPA-specific glyphs as stylistic variants is the way to go as the software capabilities catch up.

charles ellertson's picture

Peter, Brian

Peter probably has more to contribute to this thread at this point, but let me echo a few things he mentioned.

First of all, while I use phonetic symbols reasonably often, the greatest use I encounter is in languages which never had a written form -- native American languages come to mind -- and the various scholars adopted various orthographies to come up with a written form. In the older days (e.g., Ella Deloria), they used what was convenient to note with a pencil, writing in the field - i.e., diacriticals. Later, phonetic symbols were more common,and are more often used today.

Secondly, by in large the symbols used for phonetics were originally just what was available in type, specifically, foundry or Monotype, with their singly cast letters. So, a turned "a" was just that. A long s was just that. I'll speculate that over time, special sorts were created, to aid clarity and consistent usage, but I'm not really a type historian.

Dictionary use was probably the financial motivator for making whatever custom sorts were needed. But for dictionaries, italic phonetic symbols were not needed, or at least, not often needed. It was other uses of phonetic characters that brought in the need for italic, and those were quite specialized -- i.e., had little potential for recovering the financial cost of making the characters.

Thirdly, not everyone uses the IPA system. There are others.

For all these reasons, a separate font containing IPA characters is not going to solve all users problems. It may solve the dictionary publishers problem, but that is not the only need.

* * *

Linguist are just going to be limited in the number of fonts they can use. The problem is, at this point the number of fonts suitable for bookwork approaches zero.

The number of available fonts increases if not all that many phonetic symbols are needed. As I've said, for foundries that permit modification of their fonts, the end-user can add to the character set. But as the number of needed characters increases, and as the use moves toward serious linguistics, more and more precise characters are needed.

For my money, for serious work involving academic linguistics, I would be better off to fix the several non-phonetic problems of Charis-SIL for bookwork (high-resolution offset printing) than trying to draw up all the needed phonetic characters. Others, whose skill level in drawing up characters is greater, may take a different approach.

David W. Goodrich's picture

As a footnote to charles-e's point about losing kerning between fonts, I wonder whether this might be overcome with something akin to Peter Kahrel's Scripted Custom Kerning Tables for InDesign. These allow one to set up kerning tables that work cross the different font styles in a family, so that one can prevent, for example, italic-f from crashing with an upright closing parenthesis. I confess I haven't actually tried Peter's script, but am anxious to experiment as soon as I can find the time.


Peter Farago's picture

When you say you would prefer the stylistic variant approach, is that because you already work with OpenType- or AAT-savvy software?

Yes; for serious work I primarily use XeTeX for papers and Apple's Keynote for presentations, both of which support OpenType variations.

In general, would linguists and other end-users of IPA fonts have the software to handle the stylistic variants? In your experience, how would they handle these types of issues?

Linguists seem divided into two groups on this respect: on the one hand, I have encountered professors who still use SIL's symbol encoded legacy fonts because they have yet to master unicode entry. On the other hand - and I think this represents the majority of young linguists - there is a startlingly high level of typographical literacy. Nearly every publishing linguist is TeX-savvy and uses two or more linguistics-specific packages. A short ReadMe bundled with a typeface the likes of which we have been discussing, including some example XeTeX macros to enable stylistic variants, would probably be received very well.

Andreas Stötzner's picture

Peter F.: … who would appreciate your taking the trouble to work out typographic oddities like the italic two story “a”

I am happy to confirm that I take such trouble already for several years, more than others do. Andron Mega Cp. Italic has *three* different two-storey-a’s alone (among them special medievalist glyphs):

Knowing anyone trying harder -?

Andreas Stötzner's picture

Charles E.:… the number of fonts suitable for bookwork approaches zero.

I disagree. It’s *one*.

Peter Farago's picture

Andreas, have a look at the samples from IPAPhon in the Greek Characters thread to see what these symbols should look like in IPA contexts.

Andreas Stötzner's picture

A good phonetic font package is coming: :-)

Sylph's picture

Andreas Stötzner's picture

A good phonetic font package is coming: :-)

It seems this didn't get finished. Or did it?

HVB's picture

Isn't this it?

Note that the offered zip file itself contains another zip file without the .zip extension.

Andreas Stötzner's picture

> It seems this didn't get finished. Or did it?

No, it did not yet.
Still on the wish list.

Loosely connected to this: I am currently working on special phonetic characters for the DANIA transcription system. It occurs that the a/ɑ-problem gets even darker with this, since Dania is usually set in Italics only…
The whole thing may lead me to screw up this phonetic business again, neccessary as it is.

Syndicate content Syndicate content