Inflection in speech synthesis (was: Re: Some issues with the ibmtts output module of speech-dispatcher)

[Hynek Hanke]

Justin Ekis píše v Po 05. 06. 2006 v 18:52 -0400:
> You mentioned wanting more noticeable inflection. Have you tried any of 
> the other software speech options, and would you say the inflection is 
> better than some of them? Is it not as good as eloquence? That's the top 
> thing that sets eloquence above all software speech I've tried on linux 
> and I'd be disappointed if that wasn't the case here.

Hi Justin and others,

I'm interested if you or others tried Festival. Do you know it is
possible to adjust the inflection span with
	(set-pitch-range value)
in /etc/festival.conf (will soon be adjustable through Speech Dispatcher
too)? Bellow is included a description of how to test it easily.

Also I want to be sure nobody mistakes Festival voices with Flite
voices. Even though they have the same name for using the same bases
for their diphone databases, they are not the same voices at all.

If you are still not satisfied with the inflection, could you please
provide feedback on what is wrong with it (if possible, including the
voice name) ?

This is VERY important, because Festival is currently the most advanced
Free Software synthesis tool to my knowledge. Most parts of the
synthesis process, the databases etc. are not fixed by the author, but
they are written as modules in Lisp and they can be fixed/rewritten as
well as new modules can be added by third sides. Everything is
configurable. So if people are interested in Free Software and the
principles behind GNU, Linux and its technical features, the right way
now to get a better synthesizer is to support Festival by providing a
list of requested features and bug-fixes. (Another way would be
to create a whole new framework for software synthesis, similar to
Festival, but that would be a big project compared to this.)

I do not claim I currently know anyone who has time and resources to
work on it, but if a wish-list is created, someone might be found.

Thank you,
Hynek Hanke

How to test inflection with Festival
====================================

A way to test it is to run the festival binary
	$ festival
and type in
> (SayText "This is the default inflection")
> (set-pitch-range 200)
> (SayText "Hello, how are you? I'm fine, what is new? Inflection is
bigger!")

Once you are at it, you can also test the available voices. You get
the list of available voices by

> (voice-list)

and then you can switch them by issuing a command composed by the
string "(voice_", the name of the voice and the closing parenthesis.

> (voice_kal_diphone)
> (SayText "something")

or

> (voice_ked_diphone)
> (SayText "something")

and so on. For more commands, see Festival documentation. Maybe there
are more settings that can influence inflection in a significant way,
I'm CCing a Festival developer from Brailcom who can maybe say more
about them.