Using Amazon Alexa device as LMS player

[philippe_44]

Per our PM, I cannot get the skill to query the audio, but I have a Bridge whose only job is to create a virtual player and make the content accessible through a webserver.
So, I've used foobar2000 to tried that vs stream.mp3 method. It takes ~10s for playback to start with stream.mp3 method and it's immediate with the webserver bridge. I'm a bit surprised, but I've tested that enough times to be sure. It's the same whether foobar plays before playback is started on the virtual player in LMS or after.

[Paul Webster]

Quick update ... there was a mistake in my code that meant things got confused when there was more than one user of this mechanism.
That is now resolved and Philippe is able to use the hacky "skill" that I made.
He has a prototype running of an alternative to the LMS stream.mp3 - but to have a basic standalone player you can still use the stream.mp3 method.
I have had 1 additional request to try it out and have sent details to that person via an Amazon beta test invitation.

My hope is that one of the builders of the existing Alexa skills takes this on board and includes it in their stuff rather than my hack but it is there for those that ask to try it.

[castalla]

Thanks for the invite ... struggling with network/router issues a the moment, so can't try it yet.

[buck]

hmmm....Ill be watching this thread. I have been on the fence about picking up a smart speaker. My Squeezebox Boom has been sounding a little sickly as of late, so a smart speaker like the Amazon Echo is what I've been leaning towards but have been discouraged by the lack of support for a local music library. Being able to utilize LMS with Alexa would be awesome! Thanks for taking a stab at this!
Buck

[Paul Webster]

Send me a PM with your email address and I'll add you into the list of testers (when you get your Echo compatible device).

[buck]

I will do that. It probably isn't going to be in the very near future. My Boom still is good at lower volumes. Its more for when we are outside and want a little higher volume of music around a fire pit that the sound deteriorates on it. I live in the State of Minnesota. The weather here has not been friendly to backyard fires lately. :-)

Buck

[malted]

I just read the thread and it inspired me to cut corners. I just hijacked a radio station on tune-in, that alexa knows and redirected it to the logitech stream url. It works and doesn't require any alexa skill. It suffers from the rather large delay due to buffering. But beside that, it works great.

In a nutshell:

If you say play "Lucky 7 Music Radio" the echo device plays http : //ic2514.c1916.fast-serv.com/stream.m3u (I'm not allowed to post links)
So I changed to DNS entry for ic2514.c1916.fast-serv.com to point to a local lighthttp, just dropped a stream.m3u in the www-directory which contained one line: http : // lms-ip:9000/stream.mp3?player=Echo (remove blanks)

It might not be as elegant as a custom skill, but it's good enough for me.

[Paul Webster]

Yep - that is broadly the same approach but does need the user to have ability to manipulate local DNS and run a web server (although I suppose it would be possible to use LMS itself as the dummy web server by putting in a tiny plug in).
If you want to try my hack just PM me your email address.

[malted]

Yep - that is broadly the same approach but does need the user to have ability to manipulate local DNS and run a web server (although I suppose it would be possible to use LMS itself as the dummy web server by putting in a tiny plug in).
If you want to try my hack just PM me your email address.

I will.

Fiddling around with lms and the echo as a player, the latency is really a disaster. It's around 60 seconds here. I tried to reencode the lms stream.mp3 with ffmpeg to use HLS.

ffmpeg -fflags nobuffer -flags low_delay  -i http://192.168.1.78:9000/stream.mp3?player=Echo  -acodec aac -strict -2 -hls_time 1  -hls_wrap 6 -f hls stream.m3u8

I hoped to decrease the buffer in the echo device through chunking, but it doesn't really work. I'm even considing streaming wav/pcm to the echo device, because the buffer could be set to a certain size and would be filled quickly with very few wav seconds... does anybody have a idea, how to reduce latency?

[philippe_44]

I will.

Fiddling around with lms and the echo as a player, the latency is really a disaster. It's around 60 seconds here. I tried to reencode the lms stream.mp3 with ffmpeg to use HLS.

ffmpeg -fflags nobuffer -flags low_delay  -i http://192.168.1.78:9000/stream.mp3?player=Echo  -acodec aac -strict -2 -hls_time 1  -hls_wrap 6 -f hls stream.m3u8

I hoped to decrease the buffer in the echo device through chunking, but it doesn't really work. I'm even considing streaming wav/pcm to the echo device, because the buffer could be set to a certain size and would be filled quickly with very few wav seconds... does anybody have a idea, how to reduce latency?

I think the only option is to starve the player with pseudo-real-time stream. I did an Alexa version of my bridges and I discussed that with Paul. It works as well, but for me, the real showstopper is that you can’t really control it from LMS. We did not find a way for commands (play/pause/stop) to be relayed to a skill. So I can easily do the audio data ´starving’ but I don’t feel it will be a great solution anyway. In addition, in the tests I’ve done you can sent pcm to Alexa (in a wav or in an audio/L16 format) so the pseudo real-time will not be accurate. My bridge can decompress any stream on the fly to à given format amongst plain pcm, flac or mp3 so the only viable option would be mp3 at CBR and use that for throughput estimation ... not great anyway