Multi-Sound Card Squeezelite Docker Container

[uncut otter]

Really excited with my latest work on getting us off "Hey Google" and onto an HA powered voice assistant. One of the immediate needs was multi-room audio control. We have in-ceiling speakers in each room plumbed to central location, we've basically just not used them and the kids are very familiar with "hey Google play sirius xm the message" so we need a direct replacement. Enter squeezelite Multi-Room docker container. Plug 6 USB soundcards into the dockerhost, create a new player in the UI for each, set hardware volume and now Music Assistant (and by consequence HA) can send room audio to each indvidually, as a group etc etc. https://github.com/chrisuthe/squeezelite-docker If anyone needs something like that!

[hybrid fable]

How good is the voicecontrol?

[uncut otter]

Using the HA box- voice control is fine for specific use cases and needs but doesn't replace Google Home in a generic way. If you ask questions like "How long is an elephants trunk" vs "Can you turn on the lights in the living room" it's not for you. Working on using a local LLM for the back and forth which might help but I"m not deep enough in to know if that will be the watershed moment.

[uncut otter]

there is a good number of blueprints etc aroudn voice controlling Music Assistant, it's not quite to the "native" feel yet but it's pretty OK if you just use the same phrase structure

[hybrid fable]

I mean control. I have my own written commands. Play local music and albums. Wonder how good hass has become

[uncut otter]

for that HA + Music Assistant is very good.

[hybrid fable]

Is the docs then behind. Voice ControlPreview image
Home Assistant has very limited (intent) support for music/media control and there is no support for initiating playback of music by voice.

https://github.com/music-assistant/voice-support/tree/main/local-assist-blueprint

Still very primative. Dont think it will handle typos well

[uncut otter]

I don't think you're going to get much better unless you provide a local LLM and use option 3 (my plan): https://github.com/music-assistant/voice-support I'm working on which LLM will be the right size/accuracy trade off for our house

[hybrid fable]

I have a self made system that uses the levenheith (something) function to check the closes matches. And my llm is feeded with artists (album of artists) radio stations and podcasts

[uncut otter]

The Tool provided by the MA Github would let an LLM look up existing artists, radio stations etc that are avaialable in Music Assistant

so it could make suggestions, matches, etc if the tag data is good.