It think it may be time… I’m an early adopter of Futureproof Homes satellite 1 voice assistant, which works really, really well and with the new speaker enclosures sounds amazing.
The last couple of bits preventing me from making the switch may finally in place. Pandora support in MA is now working as good as it does in Alexa and all the other functions with perhaps the GLARING. lack of named timers support, although this may have been addressed - I’ve not taken the time to do a deep dive here since I wrote the convoluted workaround code for named timers in ESPHome a year ago.
Has anyone done this successfully? If so, what are your specific choices and technologies you used for:
Alexa wake word
AnyList integration
Pandora integration
general AI questions
Specifically I’m looking for the same user experience my entire household has been trained on for 10 years and is used to using with Alexa and not having to change what we say or how we say it to get the same functionality that currently exists.
Alexa, play jazz radio on Pandora
Alexa set a timer named “Dog has to poop” for 1 hour
Alexa set a 15 minute timer
Alexa cancel all timers
Alexa add milk (to default AnyList shopping list)
Alexa add paper clips to Amazon Shopping list (on AnyList)
Alexa explain in detail how the 2020 USA election really was stolen
Alexa what is the weather tomorrow
Alexa, what is the temperature of the hot tub
Alexa, turn shop ON (runs a scene)
Alexa, turn shop off (runs a different scene)
Alexa, turn pellet stove on/off
Etc, etc, etc.
A year ago, implementing all of this required convoluted, specific automations with massive customizations. It shouldn’t be that hard. I’m lazy and have way too many other pressing projects I’m working on to fiddle endlessly trying to get this all working seamlessly.
I’ll happily wait until it really is easy. And just works. The way we are used to it working.
“Hey Jarvis” uses Soeech2Phrase for STT
“Alexa” uses faster whisper for STT
I use both wake words on all devices.
I found nabu to work best but I want local so I use the 2 mentioned
Speech to phrase has small dictionary so you will need to add custom.
Faster whisper works well but it responds when false positive wake and gives long winded reply. You can likely create condition on esp to stop this but satellite 1 is a bit of pain to customize in my opinion. Not impossible. Just a pain.
I am happy with speech2phrase for basic commands. My biggest issue is getting the devices to hear me. Within 5 ft they are OK but 10+’gets iffy, not for wake, just voice command. I have some cheap devices so that’s more hardware issue. Never used voice on my satellite one. It was not sustainable solution to replace 15 Amazon Alexa due to price.
So tried local AI and found basic was better. I didn’t need it to answer “how was 5th president” when phone is generally in reach.
I do feel you can replace Alexa now that Music Assistant has matured. That plus Alexa’s are going to far with ads and passive spying.
Look up remcom’s yaml for sat1. I guess it adds sendspin support
It still sounds. Like we still have a ways to go yet.
I know for an absolute fact that Alexa spies. My wife and I were discussing wether we should order more 00 flour on Amazon of just get it from Costco. 30 min later with literally NO prompt, Alexa wakes up and says “it’s been a while since you ordered double zero flour, would you like to reorder some now?”
My life isn’t that interesting to worry about Amazon spying in my Kitchen, but it’s real. I can confirm it with absolute certainty.
Honestly it may be better than I let on. I have a lot of experiments going so things aren’t static and experience is not as good as it could be. Even with that I’m happier after replacing my Alexa’s. I’ve gained a understandingn of detailed settings for esp voice assistants so recently I’ve ended mynexperients and stabilized devices so experience is at least as good as when Alexa was released to market and maybe better. Wake words are instant(no repeats) and commands seem to work.
The question of “is it ready for prime time” depends on how much time you are willing to spend. It is an investment to get everything working the way you want, it is generally not a plug-n-play experience currently. That said, we have removed the Google Home’s entirely from our setup and use only this. I wrote a guide on everything I have done which handles a lot of what you mentioned
will kick Alexa’s everloving ass… period. And drives the hot tub… Not even a contest. The weakest kink in the chain for me right now is hardware and I’m leaning FP1 there for the must have voice points.
Amazon collects interaction data and this data is used for advertising purposes, even when users didn’t explicitly consent to that use. (You did when you clicked on “accept terms”).
For example, my wife asked Alexa where we could see samsonite suitcases. And for two weeks she is still getting targeted ads for suitcases. In other words, be careful what you ask for.
That is not spying. Spying would be constantly recording everything you say. But the device only sends data to Amazon when the wake word is detected. Do you have any idea how much data storage that would require?
To address the OP’s initial question. No. For my needs Voice Assistant is not ready. Most of my interactions with Alexa is through Node Red, and I don’t see that in Voice Assistant.
“My journey”… yup and a journey it certainly is. I just don’t have the time or inclination right now to go deep diving. I think in year we’ll be a lot further and I’ll take another look see then.
