Automations often can be triggered unexpectedly. If an automation creates sound announcement (either just by playing a sound effect, or by text-to-speech), it can be a problem if another automation is also triggered while the first one is still playing, because then the second will stop the first sound.
I couldn't find any media player capable of dynamic queues, hence this Appdaemon script. It calls HA API to generate mp3 from text, and from the length of that file it determines when will be the speaker free again. If another event comes in, it is scheduled to be played only after any earlier sound is completed. Appdaemon cannot create service callable from Home Assistant, but it can listen to events, which has the same result.
Because I use many sound effects played before TTS announcements (like a bell, a dog bark, etc.) events can have also a filename parameter beside message, to specify the filename of the sound effect.
Have this in appdeamon's config:
python_packages:
- mutagen
Create a long-lived token in Home Assistant, and copy its value either into apps.yaml, or into secrets.yaml.
tts_announce:
class: tts_announce
module: tts_announce
speech_token: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
ha_url: http://192.168.x.xxx:8123
tts_platform: google_cloud
tts_language: hu-HU
speaker: media_player.mpd
night_volume: 0.6
day_volume: 0.8
debug: True
extra_delay_if_sleeps: False
Where
| Key | Required | Description | Default |
|---|---|---|---|
| speech_token | True | a long lived token created in Home Assistant | None |
| ha_url | True | is how you access Home Assistant on the local network. It is used to access the TTS API, and also as url for the files to be played | None |
| tts_platform | True | The same platform you specify in Home Assistant configuration.yaml under tts | None |
| tts_language | True | Default language for TTS. Can specify a different for any event call | None |
| speaker | True | Id of the media_player entity to play sounds on | None |
| night_volume | No | Volume level between 0 and 1 for the night (22-05) When set to 0 or left out from apps.yaml, no volume change will take place | 0 |
| day_volume | No | Volume level between 0 and 1 for the day (05-22) When set to 0 or left out from apps.yaml, no volume change will take place | 0 |
| debug | No | Log every detail of processing in Appdaemon's log when set to True | False |
| extra_delay_if_sleeps | No | Can be set to True for Google devices to have an extra 3 secs delay when the speaker is in "off" state, because it takes about 2 secs to wake it up | False |
| media_path | No | Sound effect files local path | /local/media/ |
speech_token can also be placed in secrets.yaml with apps.yaml containing
speech_token: !secret speech_token
just don't forget to restart Appdaemon after it, as it caches secrets.yaml
TTS message started with a bell:
action:
- event: tts_announce
event_data:
filename: bells/proxima.ogg
message: "Someone is at the door"
Where it could be used with either only "message:" specified, when no effect is required, or only with "filename:" parameter when no TTS is wanted.
Optionally the delay can also be specified for example when it is a long mp3 and only the first 5 seconds requied:
action:
- event: tts_announce
event_data:
filename: effects/very_long.mp3
delay: 5
message: "And this is the announcement"
Language speech parameter can also be specified instead of the default one.
action:
- event: tts_announce
event_data:
message: "Guten morgen schönheit!"
language: de-DE
Randomized effect and message with template action:
- event: tts_announce
event_data_template:
filename: >
{% if (range(1,100) | random | int) > 50 %}
pumukli/jol_aludtal_kakaot.mp3
{% else %}
pumukli/jo_reggelt_kivanok.mp3
{% endif %}
message: >-
{{ [
"First funny message",
"Second",
"Third"
] |random }}