Re: IPA Text-to-Speech

Outo Otus wrote:

> Hi. I was wondering if there is such a thing as an IPA text-to-speech
> program, or something similar. What I mean is, a program that when you
> speak, it transcribes what you said using IPA symbols...,
I suspect one needs the human ear to capture all the subtlety and variation
of the flow of speech........Though evidently speech-recognition programs
can do an adequate job; how else does the computer on the other end of the
phone know what we're saying??  Trebor Jung must know something about
this....?

>and if you write
> words in the IPA it will speak them to you.
This must exist-- speech synthesis programs.  $$$$$$$ no doubt.
Someone once posted a URL for an amusing site that could do this in a mini
sort of way-- you typed in real or conlang words (not in IPA however), and
could select to hear them pronounced as if they were Spanish, French or
German words.  It worked very nicely for Kash, but Kash has pretty simple
phonology (almost = Spanish) . Actually IIRC the German version was best,
the French version thought-provoking, the Spanish least realistic. Go
figure.

Fascinating idea. I'd love to get hold of such a thing. Given a set of
rules, a set of phonemes, and an IPA sound file, it ought to work.
Intonation has to be the hardest part.

Way back in grad school, we played around with a speech synthesizer that
someone had built in the 1950s at U.Michigan.  Full of vacuum tubes and
about the size of a VW Bug :-))) but it worked more or less. IIRC someone
managed to get it to produce French.

>I think this type of thing
> would be useful in a number of ways, mainly for linguists, but for other
> people too. For example, if you wanted you to know you were pronouncing a
> word correctly in another language, then you would be able to check if you
> said it correctly, because you could see exactly what you said and where
> you went wrong, and it would work for every language.
For that you might want a sound-spectrograph program (I believe one exists--
again, probably $$$$$).  That would show that you were producing things the
right way (correct formants etc.), rather than sounding "close enough" but
for the wrong reasons.

Replies