Replies: 2 comments
|
That's a little odd, I also have a somewhat deep neutral-accent American english voice and in general performance seems fine. If you use openWakeWord by itself instead of via Home Assistant, is performance similar? As for pre-trained models, there seems to be a large collection of community trained models here: https://github.com/fwartner/home-assistant-wakewords-collection And yes, it is possible to run the training of a custom model locally, but it is not always straightforward to configure the environment correctly. I would start from this example notebook: https://github.com/dscripka/openWakeWord/blob/main/notebooks/automatic_model_training.ipynb |
|
Thanks for the link to the wakeword collection. I see "computer" in there so I can just install those files and I think it will meet my needs. |
Uh oh!
There was an error while loading. Please reload this page.
I'm running Home Assistant and having nothing but problems with the stock openwakeword words recognizing my deep American voice. I've been trying to build a custom wakeword using the published Google Collab but it kept timing out and failing and now I have zero credit hours left on a free account.
Is there anyone that has already generated an openwakeword for "computer" that could just share it for me to grab?
Alternatively, is there a way to generate the openwakeword locally? Maybe a tutorial on how to install all the packages and run them in a Windows subsystem for linux session?
All reactions