Long-time lurker, occasional poster here with an overall well-functioning HA instance on a N100 footing.
While I try to reduce my reliance on cloud-based IoT services, there’s a few things that are just not as neatly integrated (if at all) in HA. One such thing for me is (akin to Alexa) the choice of a default media_player entity per room, so that “Play music in the kitchen” would e.g. always start songs on the Sonos box there.
But this further pertains to any TTS content as well. I joined the hunt for open-source voice assistants quite early as well. And apart from still questionable comprehension/flexibility rate (let’s hope the HA hardware will change that 
) one annoying thing is the bad quality of the built-in speakers (in S3 Box 3 here) when we have perfectly fine speakers sitting in the room. In my simple mind, this shouldn’t me messing with on_tts_end in yaml but rather a drop-down list of “on device” vs “on preferred speaker” or a chosen entity.
My 2 cents on the matter.
Cheers io
PS: Keep up the great work, team! 
This is merely nitpicking 