MyVolumio AI Voice Assistant

Do you think that an Artificial Intelligence Voice Assistance will be cool on MyVolumio?

  • YES, go along and do it!
  • NO, I canā€™t bother less

0 voters

Dear Volumio people!

As anticipated last week, we are really happy to announce a new partnership to develop what we think is a super cool addition to MyVolumio: a Voice Controlled assistant powered by Artifical Intelligence to control Volumio!

To quote from the original Announcement made at World Summit AI 2018 in Amsterdam:

So, as you can see, there are some great news here:

  • Cloud Constable will develop an AI Voice Assistant and integrate it into MyVolumio
  • It will be available to MyVolumio users with a paid plan (as a free addition until sometimes next spring)
  • We look forward to make something which makes sense for music lovers, so the community contribution is a vital part of this new adventure.

Our idea is fairly simple: add a voice control facility to Volumioā€™s UI which will allow you to do things like ā€œVolumio, play Tom Waitā€™s last albumā€ or ā€œVolumio stopā€ or again ā€œVolumio muteā€.

Now, before committing to this project we really want to get the opinion of you guys on this:
Is this something you want to be integrated in MyVolumio?
Is this something that can make your listening experience better?
Do you have some ideas for cool things that you see this technology will enable you to do?

Let us know!

BRAINSTORMING SUM

  • No always on voice assistant (maybe selectable via a button on the UI?)
  • Evaluate the chance of integrating existing voice controls like Alexa or Siri
  • A good feature is understanding musical tastes to offer recommendations

Iā€™m going to abstain from voting in the poll as Iā€™m obviously biased, but I wanted to take this opportunity to introduce myself to the Volumio community, and further reinforce Michelangeloā€™s message. We really think that the addition of AI and voice control is going to add to the overall enjoyment of the Volumio user community.
Bringing our AI technology into the Volumio platform is also a key milestone towards our overall mission of connecting families & friends and enabling their sharing of digital content, while protecting them from cyberthreats. Our animated AI superhero, CloudGirl, will be the social conscience of our social network.
So another question for all of you to consider is whether having a customizable, animated AI persona included in the platform would be useful or valuable. Weā€™re already experimenting with creating a ā€œDJā€ persona named Viola for Volumio users, and even have some ideas about how to give her some interesting DJ skills through using machine learning technology. Of course, all of this would be customizable per user or even just turned off if preferred.
So, who wants to hear more about this concept? Any ideas whether this would be something to integrate right into the existing Volumio UI, or would it best be built as an entirely separate add on app?

I voted ā€˜Noā€™ because the idea of having something listening all the time in my house is anathema. It might be quick and useful in some circumstances but itā€™s a witness to everything that goes on within the hearing range of the device. At some point, someone will gain access to it for reasons that I donā€™t know or approve of; thatā€™s just the sloppy nature of the cloud. Note how small the sales of smart speakers are in China. Thatā€™s not lack of money, itā€™s simply that they want some part of their life to be private from the government.

I COMPLETELY agree with you here!!! I would NEVER sacrifice privacy for convenienceā€¦ Its a sad world where others seem not to care about privacy - where so many have fought for the right of Privacy!

I voted no as well but I would like to provide more context as I am clearly not the target audience for such a feature, and I think there are simply 2 camps out there for this type of stuff.

I have a personal android phone and I donā€™t use any of the baked in Google assistant or whatever they call it this month. I had to go in through dozens of settings that were hidden away in various menus to turn off all of the tracking features that just wasted my data and annoyed me. For example if I were to eat at a restaurant without searching for it on the phone, and not even pulling the phone out of my pocket, I was receiving notifications to post pictures and write reviews of the restaurant. Presumably my phone took it upon itself to monitor my location and report it to google (cloud?) and they used AI to determine I must have eaten there based on the duration I remained stationary.

Iā€™m not niave about things - Iā€™m 99.9% positive that Spotify is using an AI cloud based system to generate a profile of my music listening to generate daily mix and other dynamic playlists and it is fantastic.

So, I would like to suggest that you should revise or expand your description about this project to include more about potential features alongside the AI assistant function with a clear note that the ā€œlisteningā€ can be completely disabled. I donā€™t see any reason why all of the voice commands you noted in your initial post couldnā€™t be typed as text (youā€™ll need this for hearing impaired folks anyway). This way both camps of individuals will support this great initiative.

I voted no because I donā€™t use myVolumio (yet) and I already use Siri and Alexa. I donā€™t need another voice assistant in my house. Iā€™d rather see a good and easy to install integration of Volumio, not myVolumio, with Alexa or Siri.

Thank you guys for the very constructive replies, thatā€™s exactly the kind of debate I was hoping for.
Iā€™ll set up a short ā€œtag cloudā€ of all the points coming up in the discussion in the first post so we have an handy sum of this brainstorming.

Personally, I really want to understand if our community thinks an AI Assistant will be good or not for Volumio before committing the time and effort to make it happen. And if yes, what should it do and to which extent.

I was expecting the privacy topic (as some pointed out, this is our day and age most frightening issue) and couldnā€™t agree more with that. I have myself a lot of Voice assistant gadgets (Amazon echo and similia) and just canā€™t keep them turned on knowing they listen to everything I say.

The idea (at least the early ideas I have on my mind) is to have a voice button on the UI (therefore user selectable and not always on) to perform voice operated commands and queries.

The rationale here is that for some complex actions and queries, voice commands can be more handy and quick than touchscreens and keyboard. However, I admit that this alone is not a point strong enough to make it happen.

Appreciate also the suggestion of integrating existing voice assistants such as google and amazonā€™s rather than having our own. From a technical standpoint now that we have MyVolumio that will be fairly trivial (it wonā€™t be possible with ordinary Volumio for mere networking reasons). However, with those services we could not get more than play this or search for that.

Donā€™t forget that Voice control is not the real deal here: thereā€™s a powerful AI component and this is the hot topic. CloudConstable is using IBM Watsonā€™s AI tech (basically the state of the art of AI as of now).

This means that we have a chance here to do something tailored to what we think is going to be cool for our music consumption habits with the help of AI. It means to be able to do something that ordinary voice assistant canā€™t do, and do it in a way dedicated to music discovery and listening.

Thing is, as of now the only cool idea I had was ā€œSearch me this album of this artist of this yearā€, and this is probably not enough, so letā€™s see if we have some more ideas on the topic (aka what AI + Voice commands can be good for in our use case).

Looking at the common use cases for AI VAs, personalization remains the most frequent one. This is an overview from the BPI about where they see AI going in the music industry (musictank.co.uk/wp-content/ ā€¦ report.pdf). From p.12 onwards, thereā€™s a discussion of automated curation of playlists and music discovery. I think that this is the most appropriate use for MyVolumioā€™s intended AIVA, especially if you can get MyVolumio to be omnipresent within the music context. Having one ā€˜control surfaceā€™ that plays my music, no matter what the player/device, is something that those of us with more than one remote control would leap at.

Chris M

1 Like

FWIW, I do understand the privacy concerns. It is of course a tradeoff against convenience, and our plan isnā€™t to make using the AI Assistant mandatory by any means. However, I can say that Iā€™ve experimented with an Amazon Alexa device that uses a physical button to initiate the voice interaction, and found that it tends to be less convenient than just letting Alexa listen all the time.
I will say that by using IBM Watson instead of one of the other consumer services we will have more control over what happens to the audio stream and voice transcription data once itā€™s been processed by our service. Of course, anyone who doesnā€™t want to use the service could simply disable it, and we could even default it to disabled by default if that is the more popular choice.

I think itā€™s also worth pointing out that one of the most popular Volumio devices seems to be the Raspberry Pi which doesnā€™t have any built in microphone capability. Anybody with serious privacy concerns just wouldnā€™t add a mic to their hardware.

Does the Volumio team have plans on rolling out their own branded hardware (or partner/recommend someone elses) with a nice long range microphone array similar to a smart speaker like echo or google home?

Other than dedicated hardware it would need to be activated by a phone app which would also need to be implemented on multiple platforms ultimately. Iā€™m curious what the team has in mind.

1 Like

About privacy, have you ever heard about snips snips.ai/ . As I understand it all runs on device, no data sent

I also do not need a voice assistant. When listening to music, a remote control and a good app is much more important.
I mostly listen at a high volume. In my opinion, a voice assistant is not a practical solution. The remote control will not be replaced so quickly by AI and similar things in the future.
A great feature would be if you have buttons in the web interface, with which you can control GPIO from Raspberry. But thatā€™s the wrong forum here.

best regards

Really really nice find!

I voted no for reasons of not wanting to feel even more disconnected to the act of listening to music.
I love vinyl, but I rarely listen to vinyl anymore, I have all my music ripped to flac, I feel somewhat disconnected already, no opening of the vinyl gatefold, now slipping the black gold out of its sleeve, gently wiping down the disc etc.
Flac rips and cds as a whole have already watered down the physical aspect of music enjoyment, voice activation is not something I would use nor want.

  • I like this idea. I voted YES.
  • I donā€™t know and donā€™t use myVolumio, but I use volumio on RPI with many plugins.
  • First step can be ā€œautoplayā€ - plying similar songs after last song on queue
  • Where will be microphone? The MyVolumio gets sound from web browser or it must be connected special device? I mean device like microphone array from seeedstudio?
  • My opinion is this feature need volume control. When you enter command, you need set volume down.
  • Some one fears about always enabled microphone. There can be created hardware switch based on relay for disconecting it.
  • I didnā€™t read nothing about language support. English is ok, but try to describe why you like this song in non mother tongue.
  • Animated super AI hero avater. I donā€™t know if it is this good idea. Remember you Clippit - Office Assistant from Microsoft. I didā€™n like it :slight_smile:.
  • I will like see more integration with current AI on the marker ( Google, Amazon, Siriā€¦). Playing music is only one part of life. I want control more by voice.

Iā€™d be into a dedicated UI button to call an AI to do advanced commands but Iā€™d really love to see simple integration with Always On devices like Alexa/Google Home even if the commands may need to be more simple then what you guys could do with a native, dedicated AI assistant .

Being able to scream out Pause, Play, Stop, Clear Que, Volume +/-, play ??? on ???, Ect while iā€™m doing something and listening to music is all I really needā€¦ Seems like the most practical use-case for a hands free assistant

Maybe more of a chatbot-esque approach?

"Alexa, Talk to MyVolumio AIā€¦ " OR Press the ā€œMic Buttonā€ on the app, lets say, to call MyVolumio AIā€¦

Just some ideas, iā€™m not sure the technical side of these things but from consumer/ smarthome techy tinker stand point.

Yā€™all Rock. :nerd: :nerd:

Competition is hard
youtu.be/xiwFnGY4xkc?t=41

It looks like they all have strengths and weaknesses, none is perfect and never will be, maybe just for the simple tasks of making a music player work itā€™ll be fine but people are always going to want more, I foresee a never ending feature request list to go along with the problem list.

Here is list of commands for Google home. For this thread is important section ā€œMediaā€.

cnet.com/how-to/google-home ā€¦ ds-so-far/

Just a small update on our progress with the AI voice assistant. We just went through a productive testing and debugging session, in which we used a laptop as the control interface for the AI. For now, weā€™re using regular Volumio and a JustBoom DAC board on a RPi 3 B+ for testing. Our program runs on a laptop, and we also run the Volumio web UI in a browser on the same machine so we can back and forth between them to see how things are working. (We use the microphone built into the laptop for streaming the voice commands, and hooked the output of the DAC board to a Paradigm portable speaker.)
The good news is that mostly, this is working fairly well! We have the common commands like play, pause, mute, volume up/down, etc. working reasonably well, as long as there isnā€™t music playing at higher volume levels. In an ideal world, weā€™d consider implementing gesture recognition for mute so as to eliminate that problem before speaking a voice command, but that would require a 3D sensor also!
We also tested the search function and it generally seems to work well also.
Thereā€™s just a bit of cleanup left to do, then weā€™ll need to see how to integrate this into a Volumio or MyVolumio release, or suitable add on app, so that it can be tested out in real world conditions by some volunteer beta testers.