Limitations of Voice in general

  • 17 October 2017
  • 13 replies
  • 553 views

I will start this with a caveat that all of this is speculation because I still have to see Alexa working with Sonos, so anyone that has is encouraged to shoot it down in a hurry if it isn't valid. It was triggered by a recent conversation here about trouble with Radio play started via a voice command as compared to via a controller tap.

On reflection it occurs to me that compared to tapping a specific line item, voice commands are convenient, but not made granular as easily as the former. An example: A station that I listen to, Radio Paradise, offers users a way to pick a stream out of a few options, that works best for them. The higher 320k bitstream offer better sound quality but also needs/consumes more internet bandwidth than the also offered lower one of 128 k. But if one can use the higher one, choosing its URL in the Sonos controller and assigning the Radio Paradise name to it ensures that the controller tap on the name will always result in the 320k bitstream being selected for play.

But if Radio Paradise were to be started with a voice command, which stream would be selected? I don't know, but I doubt it will consistently be the 320k one at my place. And what if it is the 320k one at a place where the higher required bandwidth isn't available? Both such choices will lead to poor music play quality via voice command/Alexa compared to that using the aforesaid tap the controller method where the desired URL is consistently selected.

On streaming service also there is a lot of scope for these issues to surface. I don't know about Amazon, but both Apple and Google have many versions of the same song, some of which are remastered at different times and by different people and sound different from one version to the other. People familiar with jazz will know about this by their exposure to the Rudy van Gelder remastered versions of numerous jazz CDs, but this applies across the genres as far as I have seen. Using the controller tap, one can choose the desired specific version for play, but how will Alexa deal with these many versions of the same song by the same artiste? If it does so inconsistently, it will again mean inconsistent and different sound quality from how music sounds when selected using Alexa.

One place where this should not happen is when playlists have been created either in Sonos or in the service account using the desired versions for inclusion in the playlist - assuming that if the playlist is selected for play by Alexa, the versions of the songs picked for replay will all be the specific ones added to the playlist when it was made.

These are matters that may perhaps have not raised their head among Alexa users till now with Amazon hardware, but are more likely to arise with Sonos users because they are using Sonos to obtain better sound quality than what is available via Echo.

Or perhaps this is empty speculation as I warned up front! But if it isn't what are the solutions?

This topic has been closed for further comments. You can use the search bar to find a similar topic, or create a new one by clicking Create Topic at the top of the page.

13 replies

And ironically, this issue will not arise for music from local NAS, when voice command is used to start play because that content is usually filled with only the desired versions of the song. Ironically, because at this time this feature isn't available via Alexa.

This issue may force us to stay in the same boat we are in today for local NAS, for all music play; that of using the controller tap selection method for starting music play and leaving the option of voice control only for subsequent actions of pause/stop/restart and volume changes.
Using Alexa I have addressed the multi version issue by creating playlists in Amazon Music and making sure those pointed to the version I wanted. In a surprisingly large number of cases I had to upload the version I like and sometimes Amazon did not have the song at all so those also had to be uploaded. This just highlights some of the shortcomings of music services vs. local libraries.

I think that devices like the Echo Show or Echo Spot will help with the granularity question. As voice control evolves I think the screen will be used more to display choices to clarify what it is you intended. Alexa already does this with some functions, You can for example say "Alexa, show me albums by 'artist x' ", and you will get a display of album covers with a number next to each. You can then say "Alexa play 4" and the 4th album in the list will play. I can see this evolve to select specific song versions or radio streams as well and it will be a much better experience than tapping through nested menus in an app.
Userlevel 7
Badge +22
Of the radio stations I listen to about 95 times out of 100 I select a BBC station. I noticed that Alexa appeared to choose a low bit rate version of the radio stream which was noticeable on voice radio. It is bad enough, at least for me, that unless I want to quickly listen to a particular station or have my hands full I will start the stream via Sonos.

It is rather similar to the 2problem" you mentioned Kumar that Alexa can't, yet(?), play local music in the same way it can't play your favourites or radio stations etc.
I can't recall who did, but some months ago someone here made what now looks to be a very good point - that voice will just be an adjunct to the Sonos controller, and never replace it completely. There will be some things that will always be best done via the controller.

@upstatemike - so you confirm that if a playlist is built either in a streaming service or in Sonos using specific versions of a song, the voice controlled start of the list will have those versions selected for play. That is useful to know/have.

What is needed is a workaround that ensures selection of a specific radio station stream via Alexa.
It makes sense if you think about it. A playlist is a list of paths to specific files, not a list of artists and songs, so once you get the right track in the list it will always be right. This is also why you cannot export a playlist from Sonos and use it on another platform. Apart from the fact that there is no mechanism to do so, the list of file paths would not be valid on the new platform or location. A list of artists and songs does not mean anything, you have to have the path to the specific file and that is unique to where the file is stored.
Wouldn't this be resolved when they are able to deal with "favorites" via voice, which I would assume might be included in the option to control your NAS data. I'm thinking you could then have your "favorite" be the chosen bit rate station that you request, and consequently you'd be able to define which one you want. "Alexa, play favorite BBC 1" or something like that. It would be a small difference between "Alexa, play BBC1" which would grab whatever bit rate it so chose, without checking with you.

Interesting discussion. Thanks, Kumar!
A favorite, like your local library index and your Sonos playlists, are located on your Sonos players. It's not yet clear how these will be made available to the Amazon cloud and thus Alexa. Will Sonos store a copy of this local info for you in a cloud location that Amazon can access? Will there need to an as yet unidentified device on your LAN that can link directly to the Amazon cloud? It would be useful in trying to understand when Sonos is going to address these things to first know exactly how they will be implemented.
I agree that this is a general problem with voice control. The reallity is that a user will essentially need to 'program' Alexa to understand what you mean when you give a command. Bruce is right in the sense that adding a song or statuion to your favorites will resolve the issue...if Alexa/Sonos can read your favorites and knows to look there before looking elsewhere.

But I think it really goes beyond that. What if on a whim, I want to listen to Jimmy Buffet's Margaritaville. I don't listen to it regularly, but a friend requested it. Which version to pick? It's been done as a duet before, and there are liver version(s). At some point, we need to have Alexa (or Sonos) allow us to set criteria when several options fit the search criteria. For example, eliminate live versions, or always give me the original, non-radio edit versions when there options. Alternatively, perhaps the system should start learning our trends and such, kind of like how Pandora tries to do.

The good thing is that voice assistants are still in infant stages and will surely improve. Eventually, they should get fairly close to recognizing what you meant to say instead of what you actually said, kind of like what humans do.

The good thing is that voice assistants are still in infant stages and will surely improve.

Agreed. But for now those that are missing the ability to initiate NAS play via voice should not feel left out; at this time users of streaming services and radio may well need to resort to using the controller for this as well, and use voice for subsequent changes in music play status + volume.
I like the idea of search criteria settings! No live, no radio edit, no explicit, only explicit, etc. This will probably need to be implemented by the digital assistant provider though rather than by Sonos.

I also think a lot of the local stuff is already achievable using home automation platforms. Favorites and playlists can already be kicked off using Alexa voice commands to certain automation platforms that utilize an "always on" controller on your LAN that is set up to talk to the Amazon Cloud. I can also think of at least one that will probably give you the ability to play your local songs via voice control without pre-staging them into playlists or favorites.

I wonder if it would be best to think of voice control as a total ecosystem and not get too hung up about who is providing the solution to any particular piece of the puzzle? If you insist that for simplicity every feature must come directly from Sonos, or Amazon, or Google, etc. then you may be setting yourself up for disappointment when you could already be happily using an additional product that lets your Sonos system do what you want it to do today. Of course these additional products would need to be refined and user friendly enough to be practical for an average person to use so I am not advocating super techie DIY hacks as a viable option either.

I wonder if it would be best to think of voice control as a total ecosystem and not get too hung up about who is providing the solution to any particular piece of the puzzle?

Of course these additional products would need to be refined and user friendly enough to be practical for an average person to use so I am not advocating super techie DIY hacks as a viable option either.

As long as it works and is easy for the market to implement, it should not matter who the solution provider is. Your second quoted reaffirms this, I suppose and is essential if voice is to go mainstream. Or should I say more mainstream than it has? One good outcome from that would be the reduction of the gnashing of teeth all around for every app upgrade. One can hope. App upgrades may not even be required as often, a blessing to those that won't use voice a lot. But I see this taking years, not months.
Userlevel 1
Badge +1
But if it isn't what are the solutions?

As I mentioned in my feedback thread, and others have said here, the solution (at least for radio stations) is for Alexa and / or Sonos to first look for a requested radio station / stream in your favourites. Only if a match isn't found there should it do a wider search. Then all you have to do is put the required station into your favourites...
But favourites are in neither the Amazon Cloud, nor the Sonos Cloud, so therein lies the rub. For now at any rate.