Hi everyone!
I followed the recent development of year of the voice and really like where this is going! Although I’m not a big fan of external wake word recognition, I understand the benefits of this approach. Now it should be easy to implement voice satellites all other the house, if enough compute power is available (didn’t try it till now). But there is still one thing I am missing, which is actionable voice notifications. If you already know a way to achieve this, stop right here and reply with your suggestion! If not let me explain:
I am sure all of you now the great benefits of “normal” actionable notifications send to your phone (companion app). This great feature is (at least for me) only really great if you are not at home. At home no one really wants to carry there phone around, even if it happens most of the time automatically. I prefer home assistant to talk to me directly over wifi/smart speakers. A simple announcement is already possible, but answering a question would be the next step. For example “The sun seems bright today, would you like me to lower the shades, while your are watching TV?”. I am sure you can think of a lot more cases! If you still don’t get the hang of it, Mark Watt Tech did a video a view years back on achieving the same with alexa and a custom skill.
This feature would be great to check if a very specific automation should run, so that non techies don’t fear paranormal activities.
PS: If i miss any tags or used the wrong category, I am very sorry! Otherwise I hoped you enjoyed my very first post in this forum!
Plus 1 for this. I am over telling google to do things and have built my smart home to have as little input from me as possible. The only real use case I have for voice is both normal notifications and actionable ones. I actually thought it would be easier to set up actionable notifications than have a wake word. Just like a text notification, it should be as simple as sending a voice notification, opening the mike and then waiting for a set amount of time before closing the mike again. I am excited to hopefully have this one day… Hopefully soon
I could imagine that the problem lies in running the pipeline in reverse. There is simply no trigger as far as I am aware to activate listening of the assist on one specific device. Separating this in just saying something over normal notifications and then just triggering the assist “manually” via a script or automation would be way harder than it needs to be.
My use case … as I drive close to home, I open the app and select whether I only would like to disarm the alarm or disarm + other things (ie: open blinds, garage, etc.). This is all done via automation of course. I have created an actionable notification script which will be triggered when I am entering the zone ( script is fine …I am yet to test the entering detection part). It would be great if this is an actionable voice notification, particularly when driving. I am not really a fan of “Hey google …” and even so, most of the time I forget to ask google until I an already at the front of the house
Totally agreed that voice confirmation of an automation (particularly one triggered in the car when entering the home zone – so ideally integrated w/Android Auto) would be awesome.
My installation doesn’t show “voice_assistant.start_continuous” as a service that I can call. It doesn’t even show “voice_assistant” as a device or entity at all…
Perhaps, it’s because I don’t have a ESP32 listening device configured. But since (like @jigsaw) I’m looking to use this in the car (via Android Auto), it wouldn’t help me if the only way to access it is via an ESP32 device.
Don’t know if it works with android Auto but ok on my android phone.
You can send command to open Assit on phone and listening and combinate with my workaround ( active automation waiting response) .
Thanks @will35!
At first I thought it would be better to have a simple process integrated in HA for actionable notifications, but triggering the mic provides more possibilities! I would for example use a different set off speakers (better then any esp32 could be).
I also recently found StreamAssist from AlexxIT, which could make it possible to “create” Assist units out of already present devices.
With StreamAssist i found a way to do this. You can set several values as an answer what you want. And every option doing other actions.
Also i can now doing conversations with streamassist, or ESP32 combined with alexa. Means activating the mic after the end of the sentence without a new wakeword. But only if you want this. You can activate and deactivate this function like you want (speach, button or anything else)
i may will do a fork. But i can say it is possible to do.
