OK - I have an automation that does a test announcement…
actions:
- action: assist_satellite.announce
metadata: {}
data:
preannounce: true
message: "Calendar : Alarm Hillary Lunch at green man in 30 minutes"
target:
entity_id: assist_satellite.office_home_assistant_assist_satellite
In the office home assistant, the configuration for the assistant says, preferred.
Under Settings → Voice assistant, the Assist has Home Assistant as the preferred .
In that assistant, the text to speech is set to Piper.
Under Settings → Addons → Piper, in the configuration I can make whatever changes I like to the voice… even change the voice itself… and after saving and re-starting piper, and then re-running the automation… it still stays in the voice and speed it was originally.
What am I missing please?
EDIT - According to Mchk’s advice, changing the text caused it to re-evaluate and the alterations for the original voice were then acted on.
I’m not sure if it’s something with the Cori (medium) voice or something… but if I set the voice to alan_medium with a length of 2, and then run the test automation, it stays at cori at a good speed until I go into the assist and change the voice there… and then it comes out at the other voice (some voices just stay silent) but I did change it to alba (medium) and it came out with alba’s voice, at the speed I set in the piper setting (not alan)
So I got back to Piper and retain the 2 length and change the voice to cori and restart piper. The test stays on alba, so I go back to voice assist and change the voice to cori medium in there…
…and that time it changes to cori, but the length is back to as if it was 1.
It seems like it’s taking the voice from what’s set in Voice Assist, but the length scale from what’s in Piper … unless I chose Cori (Medium) in which case it’s ignoring the length scale.
What I’m trying to achieve is cori medium… but I’m just trying to slow it down a little
For a clean test, use updated text or clear the cache.
The default voice in the add-on settings will be used when calling the tts.speak action (w/o extra options for voice), while announce uses settings from the assist pipeline.