IPA text to voice/sound

For the discussion of language mechanics, grammar, vocabulary, trends, and other such linguistic topics, in english and other languages.

Moderators: gmalivuk, Moderators General, Prelates

User avatar
Temaharay
Posts: 2
Joined: Tue Sep 16, 2008 2:16 am UTC

IPA text to voice/sound

Postby Temaharay » Wed Nov 12, 2008 8:28 pm UTC

I am looking for a program that translates IPA text into a voice. Does anyone know if such a reader exists?

User avatar
Dingbats
Posts: 921
Joined: Tue Mar 20, 2007 12:46 pm UTC
Location: Sweden
Contact:

Re: IPA text to voice/sound

Postby Dingbats » Wed Nov 12, 2008 9:51 pm UTC

I'm pretty sure there is none. I've been looking around for one too, but unable to find any.

User avatar
Bobber
contains Disodium Phosphate
Posts: 1357
Joined: Wed Jan 16, 2008 2:09 pm UTC
Location: Holme, Denmark.
Contact:

Re: IPA text to voice/sound

Postby Bobber » Thu Dec 18, 2008 9:46 am UTC

I wanted to necro this thread, as I immediately became interested when I saw it.

I can imagine that making a IPA TTS engine would be the easiest possible TTS engine to create, as the pronunciation is so consistent. I mean, it is a phonetic alphabet after all.

So yeah, if anybody has come across this in the ~month since this thread was last replied to, please let us know!
I don't twist the truth, I just make it complex.
mrbaggins wrote:There are two tools in life, duct tape and WD40. If it moves and shouldn't, use the tape. If it doesn't move and should, use the WD40.

James Scott-Brown
Posts: 61
Joined: Sun Jan 04, 2009 6:37 pm UTC
Contact:

Re: IPA text to voice/sound

Postby James Scott-Brown » Thu Jan 08, 2009 8:58 pm UTC

Not sure if its exactly what you want, but the readme (http://www.cslu.ogi.edu/tts/download/da ... E-OGIsable) for OGISable seems to suggest that it can accept input formated in WorldBet (an ASCII encoding of the IPA).
- IPA="w oU 9r l d b E t" : specify pronunciation using Worldbet

You can download it from http://cslu.cse.ogi.edu/tts/download/

Also, some have apparently had success with an online service from AT&T: http://answers.yahoo.com/question/index ... 936AAGytpq

User avatar
MuToiD_MaN
Posts: 55
Joined: Tue Jan 29, 2008 11:51 pm UTC
Location: Irvine, CA

Re: IPA text to voice/sound

Postby MuToiD_MaN » Thu Mar 05, 2009 6:34 pm UTC

There is a real need for this, especially for Wikipedia browsing, as I'm too lazy to be troubled with learning all the IPA characters.
ĉiam veti sur Duke

AVbd
Posts: 69
Joined: Wed Nov 19, 2008 1:37 am UTC
Location: The United States of Australia (USA)

Re: IPA text to voice/sound

Postby AVbd » Fri Mar 06, 2009 4:44 am UTC

Of course, the IPA represents a much larger number of possible sounds than other languages, there are various diacritics for changing the sound slightly, and often the IPA symbols used in transcription don't actually correspond to their phonetic values.

User avatar
Velifer
Posts: 1132
Joined: Wed Dec 26, 2007 4:05 pm UTC
Location: 40ºN, 83ºW

Re: IPA text to voice/sound

Postby Velifer » Fri Mar 06, 2009 5:21 pm UTC

This guy got close to what you're considering. You may also be interested in Bell labs' (AT&T) Natural Voices, which I think can read SAMPA if you shake a big enough stick at them.
Time flies like an arrow, fruit flies have nothing to lose but their chains -Marx

User avatar
MuToiD_MaN
Posts: 55
Joined: Tue Jan 29, 2008 11:51 pm UTC
Location: Irvine, CA

Re: IPA text to voice/sound

Postby MuToiD_MaN » Wed Mar 11, 2009 1:42 am UTC

Velifer wrote:This guy got close to what you're considering. You may also be interested in Bell labs' (AT&T) Natural Voices, which I think can read SAMPA if you shake a big enough stick at them.


Very informative site. Hopefully that'll help the next time I need to read that moonspeak.
ĉiam veti sur Duke

Fryie
Posts: 77
Joined: Sun Jul 13, 2008 5:55 am UTC

Re: IPA text to voice/sound

Postby Fryie » Thu Mar 12, 2009 1:22 am UTC

Bobber wrote:I can imagine that making a IPA TTS engine would be the easiest possible TTS engine to create, as the pronunciation is so consistent. I mean, it is a phonetic alphabet after all.

Only IPA is a programmer's nightmare. Even in the times of Unicode there are better things to do than to use symbols you don't even know how to input. Makes transcription work an even more annoying task than it is already. And there's no actual need for it. TTS engines usually use a different set of letters/symbols for each language, so the standard alphabet plus some of the more common special characters are actually enough for the sounds of any language. Also, IPA is phonetic, while a TTS transcription should much rather be phonemic (since the exact phonetic quality can always be deduced from context).

The problem with TTS systems is not the transcription or whatever. It is that it is incredibly hard, even with diphones and stuff, to get intonation and all those other funny things correct, since those are suprasegmental properties of language that are not usually written (except for punctuation and related stuff).

User avatar
Velifer
Posts: 1132
Joined: Wed Dec 26, 2007 4:05 pm UTC
Location: 40ºN, 83ºW

Re: IPA text to voice/sound

Postby Velifer » Thu Mar 12, 2009 2:01 pm UTC

Fryie wrote:The problem with TTS systems is not the transcription or whatever. It is that it is incredibly hard...

Good. I can only hope the difficulty keeps people in Grufunkistan (all suspiciously named "Kevin") a cheaper option than customer service bots.
Time flies like an arrow, fruit flies have nothing to lose but their chains -Marx

User avatar
MuToiD_MaN
Posts: 55
Joined: Tue Jan 29, 2008 11:51 pm UTC
Location: Irvine, CA

Re: IPA text to voice/sound

Postby MuToiD_MaN » Fri Mar 13, 2009 11:22 pm UTC

Fryie wrote:Only IPA is a programmer's nightmare. Even in the times of Unicode there are better things to do than to use symbols you don't even know how to input. Makes transcription work an even more annoying task than it is already. And there's no actual need for it. TTS engines usually use a different set of letters/symbols for each language, so the standard alphabet plus some of the more common special characters are actually enough for the sounds of any language. Also, IPA is phonetic, while a TTS transcription should much rather be phonemic (since the exact phonetic quality can always be deduced from context).

The problem with TTS systems is not the transcription or whatever. It is that it is incredibly hard, even with diphones and stuff, to get intonation and all those other funny things correct, since those are suprasegmental properties of language that are not usually written (except for punctuation and related stuff).


Well I don't think anyone is going to be treating IPA as a "language," rather, it serves as a tool for pronouncing words and phrases out of context. IPA has the tools to suggest pitch changes (though not always employed). What you're saying seems to indicate that IPA has a shortfall that I don't understand. I thought that if the IPA representation were written correctly with all the proper details, it could be pronounced correctly without the need to consult context or the particular language the word is in.
ĉiam veti sur Duke

exitcode1
Posts: 2
Joined: Thu Nov 12, 2009 7:01 pm UTC

Re: IPA text to voice/sound

Postby exitcode1 » Thu Nov 12, 2009 7:36 pm UTC

meh? found some guy's blog post with an ipa tts scriptlet.

Erquint
Posts: 1
Joined: Sat Dec 17, 2016 5:26 pm UTC

Re: IPA text to voice/sound

Postby Erquint » Sat Dec 17, 2016 5:37 pm UTC

Okay, I just registered to necro this. But you are going to wanna know this:
In my experience(I mostly use Ivona TTS engines but believe this to quite possibly apply to any), all of the TTSs I tried could read both IPA and some proprietary alphabet.
The thing is, you need to wrap the IPA text in an SSML tag:

Code: Select all

<phoneme alphabet="ipa" ph="s'æmpəl t'ɛkst">Sample text</phoneme>

Where:
  1. "ipa" is the alphabet you want to use. All TTSs that comply with SSML have to support IPA but many also have their own ones. Ivona has X-SAMPA, which is simplified so that it can be input using a standard English keyboard.
  2. "sæmpəl tɛkst" is the phonetic text you want it to pronounce. Notice that I use " ' " sign to denote accent which is a working undocumented feature. On other hand, the usage of " ː " for IPA and " : " for X-SAMPA (" ː " and " : " are different symbols) for longer vowels is documented but doesn't actually work with Ivona.
  3. "Sample text" is an optional and rarely ever used argument for alternative text-based output(not using the TTS) — it's mostly a web thing.
This works both in browsers and in any readers. It only depends on the TTS engine(voice) being capable of interpreting it.
You just have to copy the code above to where you put your text to read aloud, substitute the second argument for the IPA text you need spoken and probably get rid of the third argument.

Links:
Last edited by Erquint on Sun Dec 18, 2016 2:47 am UTC, edited 4 times in total.

User avatar
Soupspoon
You have done something you shouldn't. Or are about to.
Posts: 2546
Joined: Thu Jan 28, 2016 7:00 pm UTC
Location: 53-1

Re: IPA text to voice/sound

Postby Soupspoon » Sun Dec 18, 2016 2:14 am UTC

This is probably not going to be a very useful post, on top of everyone's various necros, but I remember downloading a DOS program from some pre-web Internet shareware source1 that used the PC speaker to render TTS if supplied with phonemes (mostly alphabetical digraphs, outside of the easy consonants) seperated by hyphens (for flowing words) or spaces (between words).

It ought to be inconceivable that, a quarter of a century later, the same technology that could accomplish a sort-of-passable rendition of something like "a:\speak h-eh-l-oh, ay am ay k-oh-m-p-yu-t-uh" on an intel 386 (286?) without even a dedicated sound card has no direct or indirect descendent of the required kind to accept the far more rigorous IPA notation.

Erquit, exitcode1 and others reassure me that I don't need to start up my own development effort to fill this niche. Which is good, because I can talk the walk, but I probably can't walk the talk... ;)


1 Was it perhaps the source based at the University of Kent at Canterbury, as the faint echoes of memory suggest? A quick check reveals that if it's not totally defunct, in this era of Google, then Google isn't doing very well to be interpret my fragmented memories correctly.


Return to “Language/Linguistics”

Who is online

Users browsing this forum: No registered users and 5 guests