seeing the additions of Assist into HA and ESPhome now having availability for microphones etc are we still meant to be using Rhasspy? or is the plan to move everything over to HA ?
From what i have read HA doesnt have Wakeword yet so i assume until then we use rhasspy but is the long term goal for rhasspy to just become part of HA ? in terms of functionality?
I am curious on the future of this as well. I don’t think that HA should try to do everything. I like that voice assistant integration can be modular. I like that rhasspy itself can be modular, picking what things we want to do what parts of the voice pipeline.
But I do have the same concerns. I love rhasspy, i love the remote satellite stuff. I like the extra programming and addon’s that rhasspy can do. I have rhasspy doing things I don’t see a need to have home assistant even involved in. So I like they are still separate. But unfortunately, just like the continued deprecation of being able to do yaml configs for things in HA, not everyone’s opinion matches and things go a different direction.
There will still be a future for Rhasspy separate from HA. In general, I think HA will have a handful of streamlined ways of doing things. For example, you can use HA cloud or Whisper/Piper for STT/TTS and ESPHome will likely be the supported way of doing satellites.
Rhasspy will let you do many more things and stay up to date with different STT/TTS/intent systems. I’m hoping the cross-over will be with the Wyoming protocol, so people don’t have to completely jump out of HA if they just want to try a new STT system, etc.
That is great news!
I have 2 questions that are a little more specific
with all the developments in ai and speech. For TTS, what is the best possible option out there for strong clear TTS playback without using a cloud system but something local?
Is there any chance that rhasspy3 and the HA integration will allow rhasspy endpoints to be used as media playback devices? and with that, will there be a way to have the TTS engine in HA be rhasspy?
I am currently kind of fudging this process using a rest call service and passing in which node I want to send it to, but it would be nice if I could use the built in SAY type of stuff and pass it a rhasspy media player entity so that notifications would be agnostic if i am using rhasspy satellites or tablets or whatever.
Thanks again for all the work you have done for rhasspy and HA!
- This largely depends on hardware, but I’d obviously recommend Piper for most use cases (no GPU, lower-end hardware). Coqui TTS has also been going strong, and they recently added onnx export to their VITS models (effectively what Piper is doing).
- Yes, my plan is to actually make use of the empty Rhasspy HA integration to add STT/TTS providers for Rhasspy 2 and 3 servers. I’d also like to add support for Rhasspy as a media player and an intent recognizer 
 This has to wait until HA properly splits intent recognition and handling apart like Rhasspy has always done
assuming that will be something you will be pushing for in HA, is there a plan as to when that might occur? its imo the best solution and the modularization of everything certainly would help when putting together a system that covered everything.
In ref to the media player side i did have a working system with rhasspy and snapcast, however i had to use a seperate nodered function to mute snapcast on the satellite when the wakeword was detected.
Would be great to have some more local functions available. perhaps as a plugin If this then that style so if wakeword detected then mute anything else playing on this satellite. im sure there is a way to do it locally but i dont have the skills to do that atm.
i barely managed to get the audio to play after several days of messing with the linux audio options. they arent very explanative.
Does piper have a web interface to be able to hit it from various points yet? Or integrated into opentts?