A few general questions before i dig deeper :)

So far, Rhasspy has been a real eye-opener, having an offline VA is pretty cool i think and I enjoy building the various intent handling flows in NodeRed.

But I have a few questions before i go much deeper…

1 - Does anyone actually use Rhasspy as a VA in place of Alexa, Google etc or is it just a development/learning experience?

2 - Is there a way to block the fake messages - if you wake it up and then say jibberish or nothing, it spits out a random intent and action which NodeRed picks up and acts on - usually in a bad way like turning off the lights or the TV :slight_smile:

3 - What are the best options for mics and a simple speaker - i have the 2-mic hat and a PC powered speaker, but the speaker is not ideal as takes up a lot of space etc.?

4 - Has anyone built a working Alexa replacement - self contained, mics, speaker, printed/made case etc??

There are probably more but i don’t want to dive in too far if some of these points are just beyond the capability of an offline VA.

  1. I am currently still use them together. I still have my Alexas and GAs online. But am trying to fully switch to Rhasspy soon. Whenever I have a command or item that doesnt work with rhasspy currently I just handle it via Alexa or GA and try to add it to my intents or slots.
  2. Try the new 2.5.10 Version. You can set a minimum confidence score. It still not 100% perfect and needs some tuning. I get some false negatives but so far i have not had any false positives anymore.
  3. Fully depends on what you want to do. Basically you can try and connect any speaker. I currently am using an old passive “Rear Surround” from an old Samsung 5.1 system connected to the 2 Mic Hat and it sounds pretty nice. You could try with an old Driver you still have lying around. Like a Visaton VS-FR7/4. Alternatively, I really like the sound output of the Logitech Z120, Creative Pebbles or Edifier R12U. They can be fully powered via the RPI USB and then Sound via 3.5mm. They dont go incredibly deep of course, but the sound quality is fine enough for me in the Bathroom, Bedroom, Kitchen etc.
  4. Yes there have been some examples here in the Forum. I also recommend to look @thingyverse for cases. I also plan to do this, but first I need a new 3D Printer that can actually print in the sizes that I require for this.

Thanks, on question 2 i have just been playing here…

Its a pretty easy fix :slight_smile:

How do we update - i’m a linux noob :frowning:

Let me give you my perspective as someone who entered his first linux commend a few weeks ago and has no idea about programming.

1: Yes I actually use it. But only for simple tasks like switching on lights, several lights at once, changing their colours etc and for silly fun things (having it TTS weird answers when asked the trigger sentence). I think google is fr more sophisticated can make calendar entRies and whatnot. I think it’s clear that the software is far from being actually end-user friendly, so also for me it’sS just a hobby with a hint of usefulness.

2-That’s a common issue and you’ll find a lot of threads from the pros on this forum. Short answer: no easy fix.

3 and 4-With hint from these forums, I use these as satellites:

https://www.thingiverse.com/thing:4685595 case
https://www.raspberrypi.org/products/raspberry-pi-3-model-a-plus/ computer (~25 EURs)
https://wiki.seeedstudio.com/ReSpeaker_2_Mics_Pi_HAT/ Mic (~10EURs)
https://www.amazon.de/gp/product/B0894NY3SZ/ref=ppx_yo_dt_b_asin_title_o02_s00?ie=UTF8&psc=1 Speakers (~8 EURs)

I find the components are decent for the price tag and fit nicely into the case. I had the case 3D printed for ~10 EURs each by a guy on treatstock.com. It isn’t comparable to a Alexa -produced by multi billion dollar company- of course.

1 Like

Thanks,

I’m unfamiliar with all the Pi models, would you think the 3a case would take the 3b+ Pi maybe? :slight_smile:

I have a solution for for the fake responses, so far it seems workable, its in my link above.

You sound like me - i have Rhasspy doing menial things and giving hilarious responses :slight_smile: I’ll probably keep Alexa for some things like football scores etc but thats pretty much all we use her for now.

It was always my intent to use my voice to task action or grap short infos.
So currently we use it for smart home things like lights, shades, radio, etc.
Weather forecast, reminder, timer, calendar, trash can reminder etc.
It really fits our needs and is fully accepted by the family.

Still looking for better mic solution.

1 Like

@kuumaur Nice, which home automation software do you use?
@DaveKearley: The 3B+ has a different form factor than the 3A+, so the linked case wouldn’t fit.

1 Like

I currently use Iobroker. Until last year I have used FHEM.

Thanks for your questions, @DaveKearley :slight_smile:

Long-winded response:


Regarding question 1, it’s my goal to make Rhasspy suitable as a replacement for things people commonly do with Alexa, etc. The online assistants will always be capable of much more, but it seems overkill if you just want to toggle some lights, open your garage, and set timers.

Philosophically, I also prefer software that a user can tailor to their exact needs rather than a user having to change their behavior to match the software. This requires more investment from the user, of course. A skill store of some kind would certainly help lower the barrier to entry for some people.

Part of my motivation too with using an MIT license for Rhasspy is the hope that others will create easy to use “all-in-one” packages with good hardware and aesthetics. I don’t have the time to create or support pre-built Rhasspy boxes. If anyone wants to make and sell them, please do! I don’t need any money, but a shout-out would be great :+1:

2 Likes

Thanks for that,
I too think there is place for an offline VA for home automation tasks, Alexa etc goes further than home automation and tries to bridge the gap between the brain and the internet, i prefer the keyboard for this :slight_smile:

I think its a great project and if i can get it working (never seen a pi before 2021) then it must be good :slight_smile: Some more of your excellent videos would be good, maybe some with more detail etc?
I’ll have a go at designing my own box, there are a lot of flavours of pi and the 3b+ & 2-mic hat does not seem to have one yet.

1 Like

A couple more…

Holding me back form deployment :slight_smile: is the fact that with the 1-mic hat it seems pretty deaf, what is the recommended/best setup for input ?

Can it play sound effects - i’m thinking for stuff like timers/alarms etc?

1 - Does anyone actually use Rhasspy as a VA in place of Alexa, Google etc or is it just a development/learning experience?

In a way that’s a philosophical question. I never had any commercial voice assistant in the first place - even though I would have loved to buy one for several years now. However I never gave in to the idea.
The idea of an internet-connected eavesdropping device that is entirely under the control of some company that makes a profit out of using or selling my data is a very dystopian idea for me.
My personal conclusion from that is: It’s ok if Rhasspy is in the end less capable. I won’t sell that privacy for more features.
Also the term less capable needs to be used carefully. In the end Echo or GH are (in my opinion) not really intelligent. Those companies just have an army of people in the background that can provide the manpower to make the device appear like it’s omniscient.

2 - Is there a way to block the fake messages - if you wake it up and then say jibberish or nothing, it spits out a random intent and action which NodeRed picks up and acts on - usually in a bad way like turning off the lights or the TV :slight_smile:

Have you tried different wake word engines and also lowered the sensitivity? I work with different sensitivities in different rooms.

3 - What are the best options for mics and a simple speaker - i have the 2-mic hat and a PC powered speaker, but the speaker is not ideal as takes up a lot of space etc.?

Can’t speak for everyone here, just my personal observations. As mic I’m using a Matrix voice and find it to be very good for the job. I have made bad experiences on the other hand with the ReSpeaker 4.
And this is the speaker model I’m using: (beware - an Amazon link, considering my rant from above): https://www.amazon.de/gp/product/B079FKSGCH/ref=ppx_yo_dt_b_asin_title_o01_s00?ie=UTF8&psc=1
I’m very happy with it (wouldn’t use it for music playing though).

4 - Has anyone built a working Alexa replacement - self contained, mics, speaker, printed/made case etc??

Mine is not pretty because it doesn’t have a case, yet. But in terms of functionality I’ve taught it to do a lot for me. Depends on your use-case. So I’d consider it to be “complete” for my purposes.
You can find some skills in the “Show us” forum as well as cases.

In general I guess Rhasspy is certainly not an out of the box solution (yet). You’ll need to get involved at least in your own setup a bit, meaning some kind of programing for skills or 3D design or handyman-skills for a nice case.
The gain of it is a device that is entirely under your control and you can tweak it to do pretty much everything for you if you’re willing to invest some time.
And the beauty is - if you put some time into making something work - publish it here and others can use it - just like you can build on top of the work of others.

2 Likes

Thanks for the comprehensive reply,

Yes, my main reason for looking into Rhasspy was that I was growing less and less fond of the Amazon intrusion, also of late, Alexa has been struggling it seems, not sure if they are working on the back end or what but I’m growing tired of the repeated failures in simple requests and it seems i’m not alone. All i want to achieve is voice control of Node Red flows and Home Assistant actions - this is almost there now.

I found a good speaker (Amazon link) [https://www.amazon.co.uk/gp/product/B0894NY3SZ/ref=ppx_yo_dt_b_asin_title_o02_s00?ie=UTF8&psc=1]

I’ll be working on a box/case soon as there are none ‘off the shelf’ for my pi3b+ and 2 mic hat. I might try a different mic setup as well.

Its almost there but not quite yet - we are still getting to many false awakenings when the TV is on, i am playing with sensitivity now :wink:

This is a great point, and is exactly what I try to explain to people who have trouble understanding why Rhasspy has trouble with such a “simple” thing as playing a song/album/artist from Spotify. That army is updating band names, artist names, etc. 24/7. It’s a great service, but it benefits no one besides Amazon or Google; whereas a community-maintained list of (multi-language) pronunciations would help everyone :slight_smile:

I hope that one day installation will at least be a lot less complicated for most people. But I certainly don’t want to make Rhasspy less flexible in the process.

Long term, I think collaborating with folks like @DANBER and @sepia-assistant will improve the space of open source voice assistants. If we can figure out good ways of sharing project components and skills, all of our communities win.

Its not really ‘difficult’ to install, before 2021 i had never seen a Raspberry pi before, I now have 4 running in the home :slight_smile: However, i did find the instructions a little fragmented and tricky to follow, I had no issue getting Home Assistant up and running and likewise Forked-Daapd music server was also easy as was the Octoprint server on the 3d printer - the thing in common was that they had pretty much solid instructions for total noobs like myself :slight_smile:

So far i’m impressed with Rhasspy, I have no interest in getting it to play music, or searching the internet, all i desire is good, solid voice control of Home Assistant via Node Red - and it’s not far from that now.

2 Likes