Rhasspy Questions

#1

My experimenting with Rhasspy has be held up by a Bug where the site-id is being altered. ok quick back story, I have a Pi Zero with ReSpeaker Hat setup as a satellite with the base setup in docker on my Virtual Server. In my initial setup I found that MQTT wasn’t playing nicely with Home Assistant and I switched from Intents to Events before I figured out MQTT was an issue. When I still wasn’t getting a response from Home Assistant when I asked the temperature I looked a little closer and found that the Base was changing the site-id from the satellite to itself.

So I’m curious what direction I should take when the Bug is squashed (I’ve opened an issue on Github).

I’d like to have a vocal response when I ask for the lights to be turned on/off/colour change. I’d also like to have it not sound like an example of what the 1940’s thought robots would sound like.

I would like to integrate Rhasspy with Home Assistant and Node Red. Do I integrate both at the same time or do something like Rhasspy -> Node Red -> Home Assistant / Home Assistant -> Node Red?

I’ve seen a bit about integrating Rhasspy with Mycroft, is this something even worth considering? I dabbled with Mycroft a bit but found that the recognized range for response to the hotward was very post-pubesent male centric.

#2

I use a rest_command to post a payload to the Rhasspy TTS endpoint.
I have set Google Wavenet as TTS and I am using automations (listening to events) to generate the text to be spoken.
So only Rhasspy -> Home Assistant -> Rhasspy.

Is there a specific reason you want to use Node Red?

#3

I need to do some more reading and see some good examples.

Does Google Wavenet work with keeping things local?

I had a few flows setup when I had Snips running. One that let me stream internet radio stations, raise and lower the volume, pause, resume, and stop playback to my chromecast.
Another flow would give me the weather, provide me with travel time to work, and play the news when I said Good Morning.

I’m not so bent on keeping the Good Morning flow but I really would like to be able use or re-create the radio flow. If I could take it even one step further and play local mp3’s that would be awesome.

#4

Yes and No.

No: The text you want Google Wavenet to speak is send to google and returned as a wave file.
Yes: That wavefile is cached and played the next time you ask Rhappy to speak the same text.

For a lot of automation, there is a limited number of sentences you have as feedback, so there is a limited amount of text being sent once to google. After that, Rhasspy plays it from cache. The same goes for the Google Cloud TTS of Home Assistant itself by the way.

With automation created in Home Assistant, you can do all those things but that takes some effort indeed. Rhasspy can trigger automations via events. And automations can trigger pretty much everything, like scenes, light, media player and what more.

I suggest you start small, because most of the time people want to have all in the first run and get frustrated.
So, try a Good Morning first and see if you can get spoken feedback.
Check my post here:

#5

Thanks for the great info.

I think I’ll start with the Wavenet stuff and work towards an in house solution. What is the voice quality like for Wavenet? I tried to play their sample but it doesn’t seem to be working right now.

So you recommend using events over intents because of the ability to use automations?

Thanks for the link, I had something very similar setup for the rest_command but instead of rhasspy_speak I had TTS. I’m going to guess and say the difference is having Rhasspy generate the voice vs Home Assistant generating the voice?
I currently have my Rhasspy Base’s IP in the url: field but with the site-id bug can I change that to my satellite ip since it’s the only physical Rhsspy device? It is only a Pi Zero. Then when the bug is fixed I can change this back to the Base’s IP?

I think I’ll start with some simple "I turned on/off the " stuff.

#6

Yes, but that is just my opinion :slight_smile:

I’m going to guess and say the difference is having Rhasspy generate the voice vs Home Assistant generating the voice?

Corrrect, but the first is also being played by Rhasspy, while the latter is playing on the Home Assistant device, which is not very useful if you have it in a closet somewhere :wink:

I do not know about this site-id defect actually, but if you have Rhasspy running on the Pi that should indeed be possible. Make sure you configure it with the correct settings as a “single instance”.