Control HA with your voice without any cloud integration


I want to control Home Assistant via voice commands. I’m mainly interested in the controls of the Sonos speakers, playing good night stories for the kids, adjusting the heating or getting some information like the weather. The important thing is that I want to do without any cloud services. I had dealt with snips a while ago, but had also read about Ada. I’m mainly interested in which hardware I can realize this with. Since I use a Raspberry PI as host system, I thought about this first.

But with this solution I could only send voice commands in the room where the PI is located, i.e. the living room. So rather unsuitable if an audio book is for the kids is required.

In my wishful thinking I send my voice commands via my Smartwatch (K.I.T.T. says hello :-)), but I don’t think something like that works yet, do I?

I am thinking more of additional hardware like amazon Echo Dot / Echo Show or Google Home / Nest. But I don’t know what is possible with it. Has anyone of you already configured such a setup and can recommend it to others?

If you can recommend other devices, always welcome. If you still have installation instructions and would like to send them to me - would be great.

I would be very pleased about feedback.

1 Like

can anyone help here?

Rhasspy Just released ‘Voltron’ v2.5 which is a very complete redesign of the previous 2.4.x series, which I’m still running. Integrates with Home Assistant, quite easily.

Hello Keith,

many thanks for your reply. I am actually more interested in hardware configuration. Meaning: Which mirophone / speaker combination can be best used for my purposes? Must it necessarily be microphones connected to the PI or can I use “external” devices from Amazon or Google?

Keyword: Wife acceptance factor

Would be great if voice control could be used in every room regardless of the distance to the PI

the only solution that work flawless is get some amazon echo for each room and you can do more stuff that you need, the down side is it use a amazon cloud for alexa

Thanks for your conclusion. I think I’ll forego voice control for now.

With Rhasspy, you would need to be within voice range (typically 1 to 3 meters) of the Pi and microphone. The good news is, Rhasspy can be run on a Pi Zero W, making it smaller and less expensive than either a Pi 3B+ or Pi 4. The Rhasspy developers are already considering the possibility of an Android app, so using your smart watch/phone are possibly on the horizon.

1 Like

Maybe also an inspiration for the HA moderators and Rhasspy makers:

  • I am using HA (Companion) app for Android (
  • would be nice if this app could record the voice as Rhasspy satellite and send it to the Rhasspy server (e.g. via MQTT)
  • if I understand correctly Rhasspy is
    able to process incoming MQTT signals from a satellite

On this way you can easily talk to your HA Android app which triggers the automation on server side. A similar concept could be realized for smart watches etc.

Would that make sense?


razzo04 is working on an android mobile app which should act as satellite for Rhasspy: GitHub - razzo04/rhasspy-mobile-app: A simple mobile app for rhasspy.

A WearOS extension was also planned (as satellite for Rhasspy). If this idea works this could be an amazing enhancement for Home Assistant: Wear OS App? · Issue #4 · razzo04/rhasspy-mobile-app · GitHub

Just imagine, you can switch on the air conditioning of your car by speaking to your watch. In times of autonomous cars, it will be bit like “Knight Rider” :-D.


Looks like voice control with esphome microphone integration might be coming in the May release! -
(Voice Assistant — ESPHome

Hi Everybody!

What is the current recommended way to use voice commands with Home Assistant without any cloud integration?

The goal is to be able to control Home Assistant with voice commands just like it would work through Google Home or Amazon Echo, but no data should leave the local network at all. Basically if there would be a good USB microphone (Home Assistant is running on a Raspberry Pi 5) which can capture voice commands from a large room, Home Assistant could actively listen for some wake phrase and then it could translate those voice commands to the text inputs that Assist understands.

Is there any such working solution at the moment?