[DECtalk] question about Kurzweil 1000 voices

[Don]

On 6/16/2021 10:53 AM, Josh Kennedy wrote:
> Espeak runs not only on windows, but mac, Linux and android as well. By
> modifying its ontonation files and the formants of its klatt voices you could
> make it emulate the speaking style of any tts you wanted.

No.  All you can do with that is change the "sound" of the voice(s).
It's speaking style has more to do with how it parses the text that it
is given.

For example, I can use the same phoneme parameters from DECTalk
and drive them with the letter-to-sound rules from the NRL and
you would *GAG* on the resulting "speaking style".

Put a period after each word in a sentence and pass that to
DECtalk.  You've now changed its "speaking style".  How has
its listenability changed?

> A-while back someone
> sent me an intonations file and variant file and when I applied it, eSpeak
> sounded almost exactly like the keynote gold tts from the early 1990s.

The goal is for more natural speech.  The fact that you can make something
sound ancient isn't an asset!

My first experience with speech synthesis was a Votrax in the 70's.  I surely
don't want something that sounds like that in the 2020's!  I want something
that sees when I type "teh" instead of "the" and says what I *meant* to say.
Even if you are dealing with text generated "professionally" (like reading
a newspaper), there are still typos.  Every time you (the reader) have to
stop and ask for something to be spelled is a distraction in your consumption
of that material.

Imagine you have ALS.  You WILL be losing the ability to speak.
You WILL need a communication aid if you want to continue
interacting with those around you.  Something that will SPEAK ON
YOUR BEHALF.

Which DECtalk voice would you like to be REMEMBERED AS?  Given the
hurdles you are already facing, wouldn't you, instead, want to
sound like YOU?