Pete's Blog $13 voice assistant for Home Assistant

Blog posts by pete_c
Oh they're finally available (they kept running out of stock). Plenty of great reviews about them, and I really 'need'/want this, but was hoping someone here would give it a shot first. Really look forward to your posts!
 
Finally it was delivered. Weird that it was shipped from TX and was held at the USPS office in Chicago for 5 days.

The easy part was programming the device with Chrome. (via Linux or Windows 11).

There is an order to do this and it is a bit different from what is documented. Will post here.

Note here running HA in a Virtual Box on an Intel CPU which make VR faster than running it on an RPi.
 
Two running here now. Its the VR part that I am concentrating on. There is a tiny speaker on the device which works.

The ATOM Echo Smart Speaker Development Kit device was easy to program on Chrome in Windows 11 and or Linux Ubuntu Chrome.

Prerequisites:

1 - Home Assistant - here running
Core 2024.1.4
Supervisor 2023.12.1
Operating System 11.4
Frontend 20240104.0

HA Virtual box.jpg

Homeseer 4 running on same box. Audio (and TTS) is a mixture of HS, Ubuntu, HA and one Alexa dot now using Russound in wall tiny speakers.

HS4.jpg
SAPI speech via multiple voice fonts and instances for HS2-3-4
sapispeech.jpg

on Oracle VirtualBox in Ubuntu 22.04 on Lenova M900 Tiny with 32Gb of Ram and SSD hard drive. The box is also running Homeseer 4 (latest version), CumulusMX and a Windows 7 32bit Virtualbox for 6 SAPI voice fonts.

2 - Local Assist Pipeline - not using HA cloud here.
A - install Whisper -
B - install Piper
C - install OpenWakeWord

assist.jpg
openwalkword.jpgwhisper.jpgpiper.jpg
 
Last edited:
3 - Install ATOM Echo Smart Speaker Development Kit
USB cable from device to PC (any Windows or Linux running Chrome).
HA.jpg

Wyoming Protocol.jpg

Note this is the second time I have installed ESPHome. It automagically installed when I installed the Atom Echo.
device1.jpg
 
Last edited:
Are you saying you have Neospeech output on the Atoms? How's the VR working? Definitely would love more details/photos/videos/holograms/mindmeld dumps :)
 
Are you saying you have Neospeech output on the Atoms?

Neospeech and all of the SAPI voice fonts are for Homeseer speaker.exe running multiple time with different speech fonts connecting to Homeseer 2-3-4. I am using two VB instances of HS2 to run Homeseer Phone with two Way2Call boxes. Homeseer 3 cuz it runs great and Homeseer 4 because I am learning it still. My 15 tabletop tables all run Homeseer Touch, speaker.exe (with whatever voice font) and OmniTouch Pro (for alarm panel).

TTS from Home Assistant sounds great. VR is working fine and faster than the Amazon VR. Only thing when I enabled VR it enabled 280 VR controlled devices came up which I will be tweaking down. Will do a video of using one of the ATOM Echo Smart Speaker to turn on devices. Note this is a test configuration for a peer in Texas who went from a home with Gb internet to a home on 20 acres with DSL Internet that wants less dependencies on the Internet. He was using Google home devices in his old house.

Also testing (on order)


Purchased a few of these to test connection to current touchscreens running HSTouch/OmniProTouch
1705784764725.png
 
Last edited:
You got me thinking. I used to use Homeseer to "speak" things with TTS. I had speakers in all my rooms, each with a mini amp. Even had one for the front porch. Then I had this audio switch where I can send it ascii commands and it determines which room the voice would play in. I had a package box on the front porch, and when someone opened it, a wireless sensor in the package box would trip the Omni Pro II which would trigger Homeseer to speak. Homeseer would send commands to the audio switcher. It would redirect the audio to speak at the front porch and say "thank you for the package."

Yesterday I finished my new package box. I bought this $7 module on Amazon, you load an MP3 file on it, and when the magnetic sensor on the package box triggers it says "thank you for the package" using its tiny built-in speaker. Recharged with a solar cell on the back of the package box. Thinks have definitely changed.
 
Pete,
When you connected the Atom for the first time and connected to its AP, what address did you use to connect to it. For most ESP32 it is 192.168.4.1 but that isnt working with this.
 
First time connection to Atom is via a serial connection and Chrome running on Windows or Linux.

I have not documented that piece above.

A set up window opens up ...

1 - initially uploading new firmware to the Atom, then
2 - a set up window configuring the Wireless Access and
3 - finally the set up window to HA comes up where it automatically installs ESPHome and your Atom device.

Will redo set up and take a picture of each screen and post it above as one image with multiple pics on it.

1 - plug in the Atom to your PC with Chrome running.
2 - Go to this web site in Chrome



pic1.jpg

3 - click on connect

will continue this thread in a little bit...due to WAF
 
So I have it connected to HA. Forgot that I didn't need to connect to the device AP when using ESPHome.
Now trying to get assist to work. Then in watching a youtube video Smart Home Junkie said that you have to use https:// in order to use local control. Is that what you have found? Do you use https:// to launch your instance of HA?
 
I left everything at default and it works fine.

Do you use https:// to launch your instance of HA?

no.

hxxp://192.168.244.174:8123/lovelace/default_view

280 items came up to control here and disabled all of them then enabled just 2-3 to test and they work.
 
So not using https:// can you control devices (ei. lights) via voice control? When I try that it shows that i have to be logged in via secure to enable microphone control.
 
Back
Top