I don’t have a clue what’s going on over there at amazon, but they took a so-so home AI (if you could call it that?) and each day making it worse and worse and less functional. If i have to hear “did you mean kitchen?” ONE MORE DAMN time after i say “Alexa, Open Curtains” I’m going to lose it!! The Echos voice recognition is absolutely terrible. I speak to it slower then I would an infant — the infant would probably understand quicker actually.
Ok rant over. Now… what are my options? I only need a TTS system for giving vocal commands, and receiving vocal responses populated from hass as well. It doesn’t need to really obtain any data itself from anywhere but hass.
Anyone using such a system? If so, hows it working out for you? Pros / Cons?
I know i’ll eat crap. But aside from HA, one of the best software I’ve ever used, is VoxCommando. Unfortunately it didnt sell the HA aspects it could do. The dev and support team and community there was strong. It started out as a companion to mostly XBMC/Kodi but some people never gravitated to it. As it expanded into broader HA functionality, it never remarketed it as it needed to.
Its a windows only app and I’m not sure how much dev is going into it much anymore but the customization and ability to do many HA things in it and the voice commands and syntax being 100% local and granular was AWESOME. I plan on setting up when I replace my windows machine.
The draw back is getting a hands free voice solution to work is more difficult with VoxCommando. Specially multiple locations would essentially require multiple instances running (which isnt a problem perse) ore setting up networked Mics…etc.
There was an alexa integration with it sortof and a weird IFTTT Google integration.
I used to run a bare bones Dell Windows tablet (not sure they made any others with a crappy pinhole mic and it would pick up just fine right next to me.
IN VC you could make complex and alternate calls, recognition ratios…etc. You could have no prefix or custom prefix (no Alexa or hey google).
It had a good web API too. I used to use HA to run some API calls to VC. (im kinda redoing things after moving and upgrading but kids and $ have changed)
I’m trying to get away from these 3 main home automation systems we have to “tweak” to work with Hass. All of them have their pros and cons, but ultimately they’re all not designed to work the way we want them to with hass. I need a local solution. I’m running hassio on a Pi4 and would just like to install and mic & speakers and use the most basic of functions Command >> Hass >> Return Response.
Wow, I don’t know what’s wrong with your Alexa devices but my experience (with five of them) has never been as bad as what you described. I also have two devices with Google Assistant and they work well except for two or three instances where they worked too well. They responded to dialog spoken on the TV yet there was no invocation of their wake word. I checked the log and nothing sounded even remotely like Hey or OK Google.
Good luck finding something that works better for you. Let us know how it goes.
I had a old echo dot / gmini.
After the recent firesale I got a bunch of google devices.
Alexa definitely far excels in the recognition department for me. Hands down. The Drop in feature of Echos is really nice too.
Alexa does a better job of understanding my 3yo and immigrant wife w/ substandard english as well.
The emulated Hue that used to work was awesome too.
I got the Google Assistant integration working but its pretty clunky and far from brainless to setup.
VoxCommando was better than Alexa.
VoxCommando w/ XBMC/Kodi was amazing the minute detail you could do.
Well, come to think of it, when i use basic Amazon commands directly with my Echos, it works fine. It’s the intergrations with Hass/Nodfe red that it cant seem to understand. Although i don’t know what a 3rd party skill would have to do with it recognizing what I’m saying? But who knows
I’m using two different Node-Red nodes to integrate Home Assistant with Alexa. One performs Hue emulation (works well) and the other allows me to create custom voice commands (works but sometimes loses its connection with Alexa’s cloud services).
Look at Almond it’s a HA thing directly.
I honestly don’t understand your problem with Google Assistant though. It works flawlessly and integrates with HA perfectly well. (I am using the manual integration not Nabu Casa too)
I find Alexa actually does pretty good, you can always ask it What did I just say?
I find Google Assistant is terrible at detecting when I say “Hey Google”, but it does pretty well once it hears me.
I’m reminded of am old joke where an elderly fellow tells his doctor that his wife is hard of hearing. The doctor asks him to quantify it but the old gent doesn’t know how. So the doctor instructs him to go home and call out his wife’s name from the doorstep, if she doesn’t respond then come closer and repeat her name. Repeat this process as many times as needed until she finally answers. When she does, take note of the distance.
The white-haired gentleman rushes home, eager to carry out the experiment. To his dismay, there’s no response to any of his calls as he progressively makes his way through the house. Finally, when he’s in the very same room as her, she whirls around, red-faced, angry as hell, and blurts out “For the fifth and final time you old fool, I’m in the damn kitchen!”
So maybe the Alexa/Google troubles aren’t the fault of the listener but the speaker?
I’m using node-red-contrib-alexa-remote2 to INIT Alexa TTS and routines, and node-red-contrib-alexa-home-skill to register a node with Alexa as a device.
That’s the thing, i do ask “what did i just say” a lot to understand the problem. I’m speaking. Like. A. Child. Saying. Alexa. Open. Shades. With a standard North American accent and clear and concise words. But yet it takes that as “did you mean kitchen?” 8/10 times. I usually have to close all the windows and doors (if open), turn off any TVs or stereos on anywhere around it (even if playing at a whispers level) and once in absolute dead silence and speaking like I would to a 356 yr old man, it then picks it up right after the 3rd or 4th time.
I use alexa-remote2 but not alexa-home-skill because that requires creating an account with an unfamiliar cloud service whose reputation is unknown to me.
thank you very much. humour is always a good start for the day.
I’ve taken the whole alexa missing commands more lightly; in fairness it works reasonably well, but it still errs every now and then.
I just take it as perfectly normal female behaviour
Apart from my sole Google Home Mini (original version, not the improved Nest branded one) my Google Homes perform really well. I find that in many places in my house multiple Homes will detect my call and then they work out which is is closest to speak the reply. Actions to control HA entities have been flawless. The only issue of late has been that after performing the action (turn light or whatever on/off) there have been announcements to say that HA cloud could not be reached, despite it actually working…? That only started with HA 0.110 or 0.111, I’m not sure exactly when
It’s like Snips but still existing