[matthew-wegner]

Some other setups already in the comments, but to chime in with my own:

I gave up on 3rd party things. My automation VM includes a GUI environment, and I run the official Linux Spotify client. The only way my setup can break is if Spotify gives on Linux entirely.

Snapcast[1] transmits two streams to 7 different speaker setups:

  * Music + text to speech
  * Just text to speech

When TTS plays on the first stream, music volume is ducked for the duration. That setup is all pulseaudio junk. I could actually play any system audio to my entire house, or even provide an 3.5mm aux input near the VM host, although in practice I stick to Spotify for convenience and the ability to use the clients on any machine to control everything.

Speakers in some rooms turn on/off completely with the room, while others stay on but toggle between music and text-to-speech, to make sure I hear those notices (which are like doors opening, washer is done, etc).

My main work setup has a snapcast client, so I hear TTS events even with noise canceling headphones on. Some snapcast clients are placed on existing machines (i.e. TV computer), while a few are dedicated Raspberry Pis.

[1] https://github.com/badaix/snapcast

[tjoff]

That sounds quite neat, however, this only works for your (spotify) account, right?

The appeal of spotifyd from an end-user perspective is that anyone on the network can control it.

[matthew-wegner]

I'm pretty sure all Spotify clients also act as "Spotify Connect" targets. Anyone could use it (also I could run an Airplay target on it, but I don't bother)

[tjoff]

Only if it is logged in to the same account.