#Cannot get voice assistant functionality to work

1 messages · Page 1 of 1 (latest)

topaz idol
#

Hi all, just trying to get some sort of voice assistant function out of home assistant. I've been playing with HA for months - love how much there is to get lost in. I recently started subscribing to GPT Plus and setup the API with HA. I followed the instructions to get it working with Assist and it seems to be operational (when typing commands in the HA browser or on the HA app).

Now, I'm trying to get voice functionality to work. The browser seems to be a huge headache (Your connection to Home Assistant is not secured using HTTPS.) I've spent probably a week down that rabbit hole, it's just so much work. So I'm pivoting to my S25 Ultra with the HA app. I have my text to speech setup with Google cloud and the Neural2 voices, but I cannot get voice to work in any way. I currently have speech-to-text also set to Google Cloud? Not sure if that is the best choice, but it's either that or "none" in the drop down. My current error in Assist on the app is speech-to-text failed. I'm going to try another speech to text option if I can figure out how to integrate it.

Just wondering if anyone has any tips or tricks to get this working smoothly. The ultimate goal is to create a voice assistant device with an ESP32 (Like an Echo Dot, or Google Nest) so any help in this regard would be amazing. I have so many hours/evenings invested, I've taken a few weeks break, I just really want to get it working this time. I feel like I copy every video or guide I follow, but some error or roadblock appears no matter what I do.

exotic moat
#

Well, first get it working. There are other STT services out there like HA cloud (if you have NabuCasa subscription) or Whisper (you can install it as add-on or in Docker, but be advised that it needs beefy hardware to work smoothly).
Then the same with TTS. Piper, HA Cloud, Kokoro, anything else that is suitable.

Regarding satellite: you may go the winding road of bare INMP microphone and MAX98357 DAC on the esp32s3. But you will struggle, if the room you place it in has minimal noise or is somewhat big-ish. Use something like Voice PE, Respeaker Lite or Satellite1 for better results.

topaz idol
#

I had tried Whisper in the past and had issues - but for whatever reason, it's working great now. I run my HA server on beefy hardware as I want to experience the full thing - no set-backs. I don't know how to enable processing through my NVidia GPU (it's a n old GTX 1060 6GB) but I may give that a try now. I don't really want to pay for Nabu, it was kind of my focus for using HA anyways, getting away from subscriptions and having privacy, etc... But if there is some good value I'm totally open to paying the yearly subscription price, I just like to try to be independant of anything else as I guess most are when it comes to DIY stuff

I am able to get speak to the app now using faster-whisper. It does take quite a bit of time before it responds. I just said "Set basement lights to 50%"
~5 sec for lights to dim
~10 sec for voice response to play

pseudo lintel
#

Does anybody know a tutorial for Voice Assistant 2025.7 or 2025.8 that includes code, setup instructions, and works 100%?"

topaz idol
# exotic moat Well, first get it working. There are other STT services out there like HA cloud...

For hardware - thanks for the recommendation! I have already had the MAX98357A DAC and omni mics along with many other sensors sitting on my desk lol. I'm still looking to see what is out there for satellites but trying to keep cost down as I want one in every room - so DIY was fun + cheap. Although I may just pick up what you mentioned so I can learn the basics and getting the software side working before diving into custom hardware builds.

exotic moat
topaz idol
#

Definitely good with micro soldering (used to repair cell phones) and have the Bambu P1S so I'm all for the DIY side. I'm in IT so pretty good with figuring stuff out - but not a fan of programming - using a lot of ChatGPT for almost all of that side. I do network infrastructure now, so that is my bread and butter.

I was just looking at the respeaker lite on digikey so that doesn't look too bad. I'll check out Koala

#

Okay, right down to the shape and ability, Koala is basically was I was building lol. That's amazing thank you so much. This saves a ton of work. My only additions would be plans to integrate RF Rx and Tx modules, IR, Temp/Humidity, and motion detection (I want it to read me off my schedule, notify of any low batteries in my HA environment and other generic information to start the day each morning when I come downstairs for work)

raw jungle
#

Do you have the Wyoming integration installed?

topaz idol
#

I also ordered a couple of the respeaker boards too. Excited to try those out with everything

feral orbit
#

respeakers work just fine, on par with the voicePE, sometimes i feel the respeaker has better mic handling.