Setting the DefaultVoice for a voice without a conf file

[Didier Spaier]

Hello Olga,

On 01/12/2018 08:16, Olga Yakovleva wrote:

Didier Spaier wrote:
>> @Olga: as there is no variant, maybe put the gender in this field as found
>> in the voice.info files instead of none? this wouldn't hurt at all and would
>> give the information directly to users preferring either gender.
> 
> If I understand correctly, this field is for listing the dialect.
> For example, when I eventually get around to trying to create a
> British English voice, this field should be used to distinguish
> British and American English voices. Or should the language field
> contain a more specific language code in that case?

>From the SSIP specification, cf.
https://freebsoft.org/doc/speechd/ssip.html#Information-Retrieval-Commands

The dialect identification strings do not have well-defined meaning yet.
If no dialect is specified by the synthesizer, the value none is used.

So yes, this field is for listing the dialect. But, what is a dialect?

For the few synthesizers I have installed, the only meaningful dialects
listed by the "LIST SYNTHESIS VOICES" command are:
*** espeak-ng ***
                     NAME                 LANGUAGE                  VARIANT
            en-westindies                       en                      029
                  english                       en                       gb
              en-scottish                       en              gb-scotland
            english-north                       en              gb-x-gbclan
            english_wmids                       en              gb-x-gbcwmd
               english_rp                       en                  gb-x-rp
               english-us                       en                       us
         spanish-latin-am                       es                      419
         persian-pinglish                       fa                     Latn
           french-Belgium                       fr                       be
                   french                       fr                       fr
            armenian-west                       hy                  arevmda
                   brazil                       pt                       br
                 portugal                       pt                       pt
              vietnam_hue                       vi             vn-x-central
              vietnam_sgn                       vi               vn-x-south
                cantonese                       zh                      yue
*** pico ***
                     NAME                 LANGUAGE                  VARIANT
                 samantha                       en                    en-US
                   serena                       en                    en-GB
                  sabrina                       de                    de-DE
                   isabel                       es                    es-ES
                 virginie                       fr                    fr-FR
                   silvia                       it                    it-IT


When adding the "LIST SYNTHESIS VOICE" feature to the generic module,
by lack of information and of a better idea I chose to write there the
generic (or symbolic) voice name, found in the AddVoice lines of the
<synth name>-generic.conf file.

In  the example above, when the dialect is meaningful it designates more
or less a territory. But at least for espak-ng this information  is
also in the synthesis voice name, as e.g. "spanish-latin-am" for 419.
Additionally probably few users are aware that 419 designates the
territory Latin America, but will understand what spanish-latin-am
means. Incidentally for this reason I don't think that using a more
specific language code like es_419 would help much. 

IMO it wouldn't hurt to write the gender in this field, if no more
useful information is available to distinguish this voice from the
others for the same language.
The SSIP specification says about the LIST SYNTHESIS VOICE command:
This feature should only be used to allow the user to choose the voice.

So lest provide the information that can best help the user to choose
(and maybe just the gender would be more useful that the symbolic voice
that I chose to write in the list for the generic module). 

I'd be grateful to hear other opinions on that.

Best,

Didier