speechd-discuss
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Setting the DefaultVoice for a voice without a conf file


From: Olga Yakovleva
Subject: Setting the DefaultVoice for a voice without a conf file
Date: Sat, 1 Dec 2018 10:04:50 +0300

Hello Didier,

Didier Spaier wrote:
> I built two RHVoice packages: version 0.7.1, and pulled from git master.
> 
> No issue to build the first one, however I could get a voice list
> neither from "spd-say -o rhvoices -L" nor in the Orca preferences GUI
> Person drop down control.
> 
> Trying to build from git master gave me this output:
> src/sd_module/list_voices.cpp:29:31: error: operator '==' has no left operand
> #if ((LIBSPEECHD_MAJOR_VERSION==0) && (LIBSPEECHD_MINOR_VERSION<9))
> 
> This made me look at list_voices.cpp. I am not acquainted with the C++
> syntax, however I know that a tabulation, not a space, should separate
> the fields in the voices list.
> 
> This is fixed with the appended patch, applied from the root of the
> source archive, like this:
> patch -p0 --verbose <$startdir/set_separator_to_tab_in_voices_list.diff
> 

Speech Dispatcher used to use spaces to separate the fields, but 
switched to tabs during the development of the 0.9 version. I tried to 
support both versions by looking for libspeechd_version.h at compile 
time. I'll see why my attempt to support two versions of Speech 
Dispatcher breaks compilation like this. It worked on my machine and I 
didn't test it more carefully.

> @ Storm: RHVoice does not associate a symbolic voice like MALE2 to
> synthesis voices at all. But now you know how select the one you want ;)

I implement support for symbolic voices by sorting the voices of each 
language by name. So, for English, Alan is "MALE1" etc. Is this 
configurable by users with other synthesizers?

> @Olga: I consider providing a RHVoice package for Slint, but am unsure
> which voices I should put in it as I see in voices/README that:
> 
> 1. The Tatar voice can only be used non-commercially.
> 2. The Georgian voice can only be used by individuals for personal use.
> 
> If I have to remove these files from the package, is it enough to
> remove the subdirs Esperanto, Georgian, Natia and Spomenka from the
> source tree or should I also edit some build files (I hope not as I know
> nothing about scons).
> 

Yes, removing the language and voice directories is enough. It's "Tatar" 
and "Talgat" for the Tatar language, though. And I will add the ability 
to include/exclude languages and voices via command-line. Only I will 
probably not be able to work on it right now, but next week.

> Or can I keep that in the package and just ship the README?

Remove them for now just in case. I will contact the people who provided 
me the recordings the voices were built from and ask them to clarify 
their license, considering this use case specifically.

> 
> Similarly, should I build the package without MAGE?

No, unless you have a specific reason. Disabling Mage will make RHVoice 
less responsive: the longer the sentence, the longer it will take to 
start speaking. Mage makes the core synthesis engine GPL V3, instead of 
LGPL. That's why some people asked me to implement the ability to 
disable it.

> 
> On the other hand, a complete package is rather heavy (no less than 250M
> uncompressed) so I wouldn't mind stripping it down a bit ;)
> 

I think that I should make it easier for packagers to package the data 
for each language separately. Then the users will be able to install 
only the languages they need. Again, I will work on it, but a bit later.

Olga



reply via email to

[Prev in Thread] Current Thread [Next in Thread]