Am a noob here with mycroft.
Wanted a female voice for the markII.
Checked at ~/.local/share/mycroft/mimic3/voices and only had en_US/cmu-arctic_low and en_UK/apope_low.
Tried a mimic3 --voice ‘en_US/cmu-arctic_low’ but didn’t work ?? so downloaded the rest of the en_US voice files.
mimic3-download ‘en_US/*’
This gave me cmu-arctic_low, hifi-tts_low, ljspeech_low, m-ailabs_low, vctk_low
Tried all thru the mimic3 cmd, but only ljspeech_low worked, perhaps because it only has a default “speaker” ?
mimic3 --voice ‘en_US/ljspeech_low’ --debug ‘My hovercraft is full of eels.’
Anyways, changed the .config/mycroft/mycroft_conf file to:
“voice”: “en_US/ljspeech_low”
and rebooted and now have a female voice.
However, occasionally (perhaps when there’s an error ?), Mr Pope comes back on !
Anyone know if I have to change the voice in some other area ? Wouldn’t have thought so.
Mr Pope obviously does not like being evicted 
This is due to the TTS cache. We pre-cached some sentences with Mr. Pope in /opt/mark-ii-product/docker/files/opt/mycroft/preloaded_cache/Mimic3 and then there is the dynamically generated cache in /tmp/mycroft/cache/tts. Clearing both caches will ensure that your selected voice is always used.
Thanks synesthesiam, removed those files from the 2 areas. So far so good.
(As I removed them, looked out the window and saw Mr. Pope exiting the building with a sack labelled “stderr” on his back, and .wavs babelling "sorry I don’t understand’ “you’re not paired” “is your computer plugged in” “help” etc.
Kudos to him tho, for donating his time and expertise to TTS.
Indeed! I hope in the far future that Mr. Pope will be the voice of the Enterprise.
Quick question, does this work for Mycroft Dinkum, or had you put the Classic Core version on your Mark II?
ecs,
nope this was on dinkum, and it works good.
I did try Classic core on the mk2, but ended up with trying to sort out it’s recognition that this is a mk2 enclosure. So temp abandonded that
Does anyone have ideas about this? I think this is correct, in that the only voices that work have only one speaker. Is there perhaps a way to convert a single speaker from an existing voice into a “new” voice with just that speaker? Alternatively, how do you get the engine to recognize a specific speaker?