I HATE Alexa with a passion. What’s other TTS command/response options are there?

Third party integrations
21

I wish i was. It’s getting worse and worse every week. The only thing I haven’t tried is looking at ways to “reset” its voice recognition stuff. I’m in Canada, so were basically at the end of the stick when it comes to updates and features for our echos, but this started happening a lot more in the past few weeks. So I’m going to check that timeframe to any updates they might have sent it. If all else fails, ill try a google home mini etc, after that… i give up. I yell at this damn Alexa more then my kids - and they’re both shit disturbers!

I think people think I’m exaggerating the problem here, tomorrow ill take a video of me speaking to it slow, clear, and perfectly understandable, and it replying with “did you mean kitchen?” Which, btw, is the response she gives to any and all commands she doesn’t understand — regardless of what i say. I could say “Alexa, booga booga booga” and she’s reply with “did you mean kitchen?” So i think there’s a bigger problem then just understanding.

22

I’ll agree with you. Alexa is Frustrating. I call her bitch at least once a day I bet.

I think what irks me the most is that if I could get access to the text of her interpretation of what I said she would be doing things perfectly. But alas. Here we are.

Also I think that having her so long and knowing things once worked one way and now they don’t because Amazon changes things is also frustrating.

If I had a nice looking box to replace her with that spit out text, I would in a heart beat.

23

Try deleting all voice history. Her comprehension levels deteriorate over time on all my echo devices for some reason, but every time I delete all voice history - it vastly improves! I do it every couple of months.

Android Alexa app - Settings - Account Settings - History - in the drop-down menu choose ‘All History’ and then push the delete button.

1 Like
24

I have pretty much the same with all of mine, she slowly goes deaf over time! Two of mine are original models, so thats probably 3/4’s through its life by now, so understandable I guess :smiley:

1 Like
25

Wow my alexa is nothing like that. I find that recognition works better when you speak more naturally than w/ an abnormal cadence where u slow down thinking that would be better but its not.

1 Like
26

Yes, absolutely correct. When they trained their STT AI, they trained it on people speaking naturally, not in a slow way.

27

Did you mean kitchen?

Sorry, couldn’t resist.

Again, not normal behavior; something is wrong with your system. The few very few requests for clarification I have received were due to genuine ambiguity (i.e. several devices having the word ‘fan’ in their name).

It may not be entirely the fault of Alexa’s voice recognition but the middleman cloud service you’re using.

1 Like
28

Do not speak slowly to Alexa, that is much harder to understand.

I have found the best success (in a noisy room with kids playing and music going) is speaking more quickly, the opposite of what you are doing. The buffer can be quite small and if you speak too slowly she won’t hear the last part of the sentence.

Seriously, try speaking just a bit faster than a normal conversational speed.

1 Like
29

What about hardware for rhasspy? Is there anything nice looking and small? Anyone have a guide?

30

My rhasspy is running on a Pi3 B as the master and Pi Zero Ws for the satellites. Performance is very good on that master and quite good on the satellites. Pi 4 is even better in performance I heard. As MICs you could use the same as for Snips, Seeed Hats, PSeye camera, some use a Jabra…

You can find out more on mine and others setups on rhasspy.org. I find it even better than Snips so far as I don’t need any webinterface and all configuration is done easily through a webinterface after fast first install. Documentation is also awesome. [/end of advertisement :smiley:]

31

If you have a server already, what’s the best hardware combo for this?

32

Depends on what kind of server you have :smiley:

You can also run a rhasspy master in a VM on your server and have Pi Zero W satellites detecting your voice. Check out the docs, they are quite extensive

https://rhasspy.readthedocs.io/

33

This just happened and this thread immediately came to mind. My 3.5 yo literally just talked w/ his entire mouth inside a small 1x1 MegaBlock (big Lego) and say Alexa what time is it. Could barely hear/understand him and sure enough she responded 420pm.

34

@ransanti was actually on the right path.

So here’s what I found out.

I did absolutely everything I could find online to fix this “understanding problem”; everything from cleaning mics, relocating, even tilting — nothing worked.

I called amazon to see if it was malfunctioning, and had a super interesting chat with a guy over there. So this is what he told me:

“The echo devices learn your speech patterns more and more as you speak to them; you’ll even occasionally hear it ask ‘did you have to ask for this more then once?’ Or ‘did I do what you expected?’ — This is Alexa learning. But what happens over time, depending on its location in your home and it’s surroundings is, it will start incorporating the background noise into your speech patterns.

So for example, if you have your echo device close to a TV or Stereo speaker, the echo will hear you say a command, and also hear whatever is making noise as well and combine them into the speech algorithm (if it hears it frequently enough when saying that command). Even barking dogs, loud children, or even loud traffic outside a window you may have your echo sitting on will pick up those sounds.

So if whenever you’ve said the command “Alexa, turn on living room lights” and it’s heard a repetitive and frequent sound along with that command, it’ll start associating it with your speech pattern over a long period of time.

The problem you might be having is you’re now going out of your way to make it silent and you’re speaking slower then usual to give the command; your echos are looking for those background sounds because they’ve been included so many times over the years.”

He went on a bit more, but you get the idea. I had my living room echo about 5 feet from a tv which is always on. Not loud mind you, but loud enough to hear. I haven’t reset anything in the entire time I’ve owned these echos. We quickly reset the voice prints and other learned algorithms, and boom, she understands me clear as day now no matter what I ask. It was a day and night difference from before. We did this 2 days ago and I haven’t had to repeat myself once yet.

So, if you’re having trouble with your echos understanding you, this might be the solution for you. But he did say it’s somewhat rare to get this bad. Alexa is designed to ignore a lot of things, including background noise during commands. It was more of a fluke that a lot of the exact same sounds were playing when I said things over the years, which lead to it attributing them to my speech pattern.

But hey, problem solved now for me. I’m happy again.

1 Like
35

I have too many hubs for my own good which is why I am deeply interested in getting as much as I can into hassio. If SmartThings sees your curtains, Harmony Hub sees your curtains and Yonami sees your curtains, Alexa will “discover” three different “devices” for the exact same curtain.

Something as simple as “Turn on the bedroom light” confuses Alexa when she see’s 3 different “bedroom light” devices. You might want to check how many instances of your devices have been discovered by Alexa. If you have more than one, go to Alexa on the web (not in the app) and physically delete them. Disabling has proven hit or miss for me but deleting has improved performance 100%.

PS - I frequently delete my voice recordings as I don’t really buy in to the voice training suggestion. I notice no diminished performance when giving commands from a blank slate.

Good luck!

36

As a rule of thumb I never keep the default names Alexa assigns to new devices (partly for that reason). If I added curtains, I’ll rename each that Alexa discovered differently, and without the actual word “curtains” in them. So, “Tuya window covers 1, 2, 3”etc.

But also, I don’t have many devices actually in Alexa. I primarily use Alexa just as a voice command platform to run flows in my HASS / node red. I have some lights assigned within Alexa as a backup if HASS restarts or is down that moment I need lights on, but other then that, HASS manages everything.

37

Also using Google Assistant here, and I must say - It’s pretty good on recognizing commands, like “Turn of all the lights” all by itself.
But I’ve also made a couple more things:

  1. Integrated HA with GA, so all my devices would be recognized by google, even when they are working offline.
  2. Integrated 3rd party service with GA - now I can create custom scripts on my PC and run them via GA
38

A little to much “Sharing” with 3rd parties for me personally. But it does make some things more convenient

39

You always can use something like Almond for voice commands, but I can’t say anything, neither good, nor bad, about it - haven’t had an opportunity to get my hands around it.
Also, you can create script based virtual switches in HA, so you wouldn’t have to share to second 3rd party system.
I’m using triggercmd, btw. It’s just really easy to configure.
And I’m pretty confident, that Google will be able to place protection for my account, and I’m also using 2FA everywhere.
So, yeah, sharing with 2 SPs, but with the ones I trust. And not using, let’s say, services like eWelink or Xiaomi servers to manage my devices.
And, dah, I’m considering the whole thing as pretty safe, but not the safest possible.

40

Dont get me wrong, I’m not looking for a 100% localized voice/response system — that would be too complicated for my needs. My main “no-nos” are Chinese servers like Xiaomi. Not only do they not work all the time, but honestly, i just dont trust them.

I have addressed my Alexa issue i was originally having with not understanding me; but i still run into issues from time to time with giving alexa a command, which goes through the node-red skill, then triggers a flow. Ill look into the options you suggested though, see what i see. Thanks

1 Like