CUSTOM wake words! Train your own words to control Home Assistant.

The next video in my series of local voice control in Home Assistant. Training your own wake words to respond via your voice control satellite devices.
Support this channel by becoming a member:
/ @mostlychris
Discord: / discord
Links:
- Wake Word Training: www.home-assistant.io/voice_c...
- openWakeWord project: github.com/dscripka/openWakeWord
If you would like to support me:
Buy me a beverage: ko-fi.com/mostlychris
Become a patron: / mostlychris
Products I reference in my videos (contains affiliate links)
www.mostlychris.com/my-smart-...
My Website: www.mostlychris.com
DISCLAIMER: Some of the links above take you to affiliate sites that may or may not pay a small commission to me. It doesn't increase the cost to you, but it does help support me in making these videos.
Want to send me something? Send it here!
Mostlychris
24165 IH-10 West
STE 217 Box 164
San Antonio, TX 78257
#HomeAssistant
#HomeAutomation
#SmartHome
#localcontrol

[@synesthesiam]

Great video, Chris! I'm Mike "The Voice" from Nabu Casa. Thanks for making these tutorials for people to get started with wake words :)

[@manfredzimmer1199]

Good Job 👍 thank you 👍

[@estusflask982]

Thanks!

[@konkoboaxel8887]

Great video, is it possible to create a voice keyword in my native language to execute commands?

[@gerryf7015]

Nice Chris. You are leading the way on these voice tutorials. Have you tried using a conference call mic/speaker as a better frond end?Anker S330. Worried about the M5 in a noisy environment (conversations, background music or TV) and the M5 speaker is useless ( in those environments).

[@edbouhl3100]

I’m also very interested in conference phone opportunities! Lots of used analog and SIP conference phones out there. A conference phone with an active IP connection to Home Assistant and a wake word is basically a smart speaker with enterprise grade microphones and built in speech enhancement.

[@gtabrpaulo]

I wonder if they will eventually enable gpu processing for stt to make this usable without using nabucasa. Currently its not accurate enough and takes too long to process (roughly 6s using base-int8 and N95 CPU).

[@edbouhl3100]

Or tap the potential of the Coral TPU, even better than GPU for some tasks.

[@Yuriel1981]

Could I run the samba share on my truenas server and use that to host the open wakeword directory so I dont have use resources on my regular desktop? I can connect it using Truenas scales inherent Samba service and create a directory on my software pool. This puts my HA and the host directory on the same PC as my HA is in a docker container with connection to the host through a bridge virt io ethernet port. I wouldnt see much of an issue unless there may be breaking issues if i get an occasional system unhealthy unprivlaged error with my supervisor. Its a simple docker restart. And it only happens occasionally with an update for some reason. But thats what i get for running supervised HA in VM lol.

[@koevoet7288]

Can you do a video on making a wakeword without google collab?

[@Perseids64]

I get an error when I try to 'play' the wake word on the colab site.When I click "Run Anyway" it says:
NameError Traceback (most recent call last)
in ()
44 )
45
---> 46 text_to_speech(target_word)
47 Audio("test_generation.wav", autoplay=True)
in text_to_speech(text)
36
37 def text_to_speech(text):
---> 38 generate_samples(text = text,
39 max_samples=1,
40 length_scales=[1.1],
NameError: name 'generate_samples' is not defined

[@craigritchie8470]

That’s what I’m getting as well.