Edit: It is, see https://mastodon.social/@danvolch...

Per https://www.home-assistant.io/voice_control/about_wake_word/#the-approach, audio is always being recorded and wake word detection is offloaded to stronger hardware.

Is there an option to enable wake word detection on device/only record audio after?

Is the hardware in the preview edition powerful enough to handle that?

If not, is this planned at some point?

Fully local processing is great for privacy - this would be a nice additional layer :)

Ty!

Like 20 December at 6:49 | Wall-to-wall | Open on mastodon.social

4 comments

@danvolchek @homeassistant I'm pretty sure that document is out of date and they're doing the wake word processing on the device itself now

20 December at 6:59 | Open on en.osm.town

Dan 🌻

@thibaultmol @homeassistant Oh - I think you're right.

The "Which wake words can I select?" FAQ describes an "on-device wake word engine called microWakeWord".

Thanks, missed that :)

20 December at 7:03 | Open on mastodon.social

shine

@danvolchek @thibaultmol @homeassistant It works really well too. I am using it with I just found this on ESP32S3 Box (https://a.aliexpress.com/_EGJE1X6) and the wake word detection is great :)

20 December at 22:23 | Open on gts.churanova.eu

@danvolchek @homeassistant if this is ever the case I hope this approach stays as an option. I actually have use cases for both

20 December at 10:02 | Open on masto.es