[no subject]

[DD]

Hi!

cxxpiper is limping along nicely, I hope to finish up my first cut
this week or next and put something out to any interested individulas who'd 
like to take a
first looook.

I got stuck for a while on the following:

This is from a newly started speechd using spd-say -o espeak hi and
shows values of SPDMsgSettings msg_settings at the entry to
module_speak_sync Note that type is 1 (male1) and not -1 unspecified,
as I would guess it should be.  Sorry if I haven't found doc for this,
but does anyone know why voice type deffaults to malle1 even if spd-say
does not specify -t male1 ?

rate: 0
pitch 0
pitch range 0
volume 100
voice name (null)
language en-us
variant (null)
voice type 1


To me, it's viscerally responsive (TM)) and the multi-speaker models
like "arctic" and "clean100" have "instant" speaker id switching per
query using spd-say -y and/or -t.  I was afraid for a bit that changing 
speakers would incur
overhead but it doesn't seem to.

Piper/onnx notion of speaker is hidden from the user (spd-say) by
expanding the modelname so that it has an integer index and speaker name
from the speaker_id_map of the model's json config, eg:

clean100~160~911

The "names", here 911, from the json files for both arctic and clean100
are pretty random looking, but they are what's in the files I have
downloaded and work to provide unique voice names for spd-say -L and -y/.

Thanks,
Derek