Autovoice

[coconutxyz]

Lol I need time to digest, I don't understand what you wrote at all..

[Desmanto]

Sorry to be confusing. My flow begin with some database initialization, which will be more confusing if I show again here. I have cut short some of the unnecessary code already before showing it here. Maybe you can try to apply one at a time, then fix it when you got error. I also have the 3 steps done 1 by 1. When testing, I got error, and add another expression, and another again. When I find the pattern that work, then I stop adding. Maybe you don't need those 3 at all, so just pick the ones you need.

[Lucy]

Ive been busy with work and christmas so all ive done is tried hotword one more time. I cant get it working good enough still. Like it works... but if sensitivity is below 50 it doesnt seem to pick up voice, even if i shout. 50 and over, it responds to all voices it hears. If it is responding to all, when i open the flow, it is red in working state and cannot be stopped unless i reboot the service.
I have noticed another tasker plugin "tasker now" i think its called. Ive had a look but its a constant refill of $ to use. How would i go about intercepting google assistant through intent or whatever? I remember reading over a post about it but was confused

[Desmanto]

You must enable "Stop listening after one hotword", if you wish to use input dialog after the first hotword. Because Hotword plugin will lock the mic usage and Autoamagic will be stuck since It can't acquire the mic access. At the end of the flow, use action start hotword again, so the hotword will be enabled again.

To sum up :
1. hotword enabled
2. Speak the hotword
3. Flow triggered, hotword disabled
4. Automagic input speech
5. Do whatever you want
6. At the end of the flow, enable hotword again.

[Lucy]

Ugh... that sux. I dont really see a relevant use for it then.
Cant leave it on cause it locks the mic, if i set it to "stop after one hotword" then at some point i have to click a button to re enable it. And as for enabling/disabling flow, i dont really see the difference. I honestly cannot see a better solution than autovoice unless i can directly intercept the assistant
(Of course im referring to a fully automated flow, which is what im after)
Oh and another thing with hotword... if it is set to stop after hotword but it activates because it thought it heard a similar word... then i have to manually again activate it. I need a much better and more reliable solution to work with my assistant that i heavily rely on

[coconutxyz]

i'll chime in and post my experience, so far it has been working fine for me @ around 27-35% , audio gain is default @ 18..very very little accidental trigger by someone else and i could trigger the hotword then continually speak to input speech no interface without having to disable the hotword. In fact the hotword is turned on 24/7 except when i am sleeping.

some hijacking, @desmanto, would need your help on this lol

viewtopic.php?f=5&t=8504

[Lucy]

Ugh.. well you got real lucky.. if i have audio gain that low, i can scream at the phone and it still does not respond and i cant leave it running all the time unless i change the whole way i use my phone so either way im screwed but hey... its ok. Its just a physical possession not real life... i will get over it

[Desmanto]

@Lucy : Yes, the problem is with the mic usage lock. This is introduced since Android pie 9.0. You have then to use another flow (App task started) to detect other app and disable the hotword service when lets say whatsapp is getting called.

As for the "Stop after one hotword", you don't have to reenable the hotword plugin service manually. At the end of the flow, use action - Plugin - Start HotwordPlugin. This is what I mean at the point 6 above. You also need to reroute the error part where it doesn't recognize a proper command back to this action Start Hotword Plugin. This ensure any success or failure in the input speech will always ended up re-enable the service. Enabling/disabling the flow doesn't change hotword service, it should be done from the action Plugin instead.

I have yet testing hotword in full 24 hours usage. I also don't use Gass "OK google" at all, since I don't need a fully voice command environment yet. I usually trigger the first input speech directly from watch instead. But you can maybe only enable the hotword service only during certain time or under certain occassion (by using other flow/trigger)

For the hotword trigger detection accuracy, it really depends on the phone, the mic, typical environment noise and the way we speak. Fine tuning is required to find the best match for 99% of the condition.

@coconutxyz : I have answered there.

[Lucy]

Well well well... as usual, look who steps up with an informative post to my request. That was a lot clearer and made more sense. Thank you. I will play around a bit. Thanks heaps guys

[coconutxyz]

not sure if you are still interested, but i came across an app call saiy and it works quite well but the development has stopped

https://forum.xda-developers.com/showth ... st77208323