Today's fun link is an enhancement to #Piper, a great no-cloud #selfhosted low-resource text to speech system. (It is used by #HomeAssistant, but you can run it trivially from the command line even on a Raspberry Pi.)
Want to make your #snarkhome a little tiny bit more terrifying? :rick: Try randomly changing out the voice for #GLaDOS! github.com/dnhkng/GlaDOS/tree/main/models
This is a little complicated (because HA) so I'm skipping a few of the finer details. Download the glados.* files, rename the json (glados.onnx.json
not onYx
) and then edit the json file (not the onnx.) Just use a text editor to set dataset to glados
and quality to medium
. Copy both files to /share/piper
on Home Assistant and restart the Piper add-on. (You may have to create the piper
directory.) Once the Piper logs show that it is running, reload the Wyoming Piper integration (or restart HA.)
That is it! To use it, go to settings and create a new Assistant with Piper TTS and the "American English" glados
voice. Experiment with the test button to say things in your browser.
Now that you have installed GLaDOS, it might be time to edit your notification automations. My #snarkhome uses templates to flip a coin. (This hacky example comes from the morning briefing, which occasionally just yells "Goooood Morning Vietnam!" at me. You will probably want to call the TTS service with the new voice assistant instead.)
choose:
- conditions:
- condition: template
value_template: "{{ range(0,60) | random == 5 }}"
enabled: true
sequence:
- event: notify_esp
event_data:
chime: media/gmv-morning.mp3
default:
- Your normal actions go here
(inspired by @EverBeyondReach)