#"take control" VPE and wakeword doesn't work
1 messages · Page 1 of 1 (latest)
You may have to reconfigure the device in home assistant after taking control as this will add an encryption key by default. check on your devices and services page to see if its popped up needing to be reconfigured.
it did pop up with a thing to add its encryption key which i did add
everything about the device is working outside the wake word
If you view the live logs through the builder tool, Is there anything that shows when you attempt to use the WW?
Nothing pops up
I get logs when I change the wake word but nothing when I attempt to say it lol
So, There was some issues that came up with MWW because one of the dependencies had some stuff changed. There was a fix merged, I am not sure if its related though.
You could try installing the Dev version of the ESPHome Builder (you can have both installed at the same time) and then try building and installing with that which will have the latest merges. The fix is also scheduled to be included in the next version of the the beta cycle and therefore hopefully in the stable release in a couple of days.
At the very least running dev will give you some more logging info by default.
ah ok cool ill do that now and see what happens
thanks
sounds like i just got unlucky with timing lol
Yeah a little bit, It happens sometimes when taking control because it pulls the current dev config and you are building it against whatever version you have installed (usually stable).
I tend to keep my production VPE's on stock precompiled firmware to avoid any of these random issues.
external_components: [source /data/packages/5444128b/home-assistant-voice.yaml:1685]
- source:
type: git
url: https://github.com/esphome/home-assistant-voice-pe
ref: dev
components:
- voice_kit
refresh: 0s
- source:
type: git
url: https://github.com/kahrendt/esphome
ref: 7a6cf5c8472b7e2fa18ee0fc314f66a80d249e32
components:
- const
- media_source
- sendspin
-
could not set 'core.repositoryformatversion' to '0'.
source:
type: git
url: https://github.com/esphome/esphome```
this is on dev
Try "clean all files" and try again.
clearing data might have fixed this actually
btw while im here is there a way to have the build pull the latest version of my package from github without having to clean all files?
There's trouble now with esp-nn library
You either have to use custom PR for microwakeword, or freeze the esp-nn lib version.
Did THIS not fix it for dev at least?
Yeah this PR.
i really picked the wrong time to do this lol
Also there were reports on VAD not working properly in that PR.
Pretty much, tbh unless you are desperate to do it now then just flash back to stock with the web tool and have another go once it settles down.
Never rains but it pours...
yeah just wanted to add a wake word and change some sounds so not too desperate lol. i guess when the stock firmware gets a new update thats a sign things have stabilized?
You can add one line to your config to mitigate it.
Life will be a bunch easier when THIS gets its turn for some dev time and gets finished.
framework:
type: esp-idf
version: recommended
components:
- espressif/esp-nn==1.1.2 # this to add
This wouldn't mitigate current issue.
Also I think this would be pretty hard on the ESP 🙂
yeah hopefully it gets added eventually although it doesn't look like much has happened for a while and i might as well get some experience messing around with esp stuff even if its only messing with a couple of lines
No but it would mean people don't have to take control in the first place.
Kevin mentioned back in Janurary that he has a version of it that is semi-working but needs finishing I think. But its just waiting for time and him to pick it back up again.
Just a matter of a finite amount of time. Sendspin did a good job of using up all the dev time.
tbf Spendspin was well worth that time its actually so insane an open source protocol works with so many random devices so well and having it on vpe is super neat
Oh Absolutely, No shade on it at all, Its just sucked up a bunch of time which was worth it but some planned things probably did get a little delayed as a result.
Yeah i have noticed that voice assistant stuff has kinda grinded to a halt tbf its not the year of the voice anymore but hoping for some more crazy cool ideas to be around the corner
There's a few things happening but yeah there hasn't been anything big for a while.
this is the log on dev ```[23:27:40.126][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:40.160][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:40.190][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:40.224][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:47.731][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:47.742][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
[23:27:47.775][D][micro_wake_word:321]: Wake word model predicts 'Okay Nabu', but VAD model doesn't.
ill try adding this