Hello all, I recently found out about the esphome-supported voice assistant firmware for the ESP32-S3-BOX unit (which I bought last year and used a bit with willow). Since yesterday I have been trying to get it to work, without success. I’m running HA 2024.5.2, and I used Ready-Made Projects — ESPHome to load the firmware onto the unit. Right now the screen shows a smiling gray house, so I believe that means that it’s connected to my HA instance (yay!). However, when I speak the wake word absolutely nothing happens. I can watch the logs and confirm that no activity takes place.
I have seen a few posts from others describing a similar problem, but most of them either reloaded and cured the problem, or the thread went dead. I have reinstalled the firmware 4 or 5 times now, so I don’t think the 7th time is going to be “the one” or anything.
Has anyone else encountered this problem? I have a BOX3 on the way, so hopefully that one will work, but I’d like to have this one working as well since I already have it.
I can’t speak for James but i’m having a similar issue with my device.
Maybe once or twice after the device reboots it will wake with ‘okay nabu’ but after that it seems totally deaf to me. It’s acting like one of my kids when they were young
Additionally, it might also be randomly waking based on some sound inperceptible to the human ear. Several times during the night i’ll hear it wake up & spout some sentence that i can’t make out since I & anyone else are in another room. Weird stuff.
I have paused my esp voice assistant experimentation for the time being until i can afford to spend more time invesetigating.
Yes, I had an assistant configured from my previous work with willow. Below the assistant, the page says 46 entities exposed 1 assist device which implies that HA knows about the existence of the S3 Box.
I received my ESP32-S3-BOX3 in the mail yesterday, so today I tried to follow the same procedure for setting it up. After adopting it into my HA install, I’m stuck in exactly the same spot. It boots up and displays the gray smiley face, but it does not respond to any wake word.
I wondered if I might be mispronouncing the wake word, so I changed the configuration on the S3-Box to use a different wakeword model (hey_jarvis). That did not alter the behavior of the box, it is still completely unresponsive to either wake word.
I just recently added an ESP32-S3-BOX-3 to my Home Assistant server and have experienced the same behavior others describe: it connects to my server, but it only occasionally responds to the wake word.
By occasionally, it’s about 1 in 20 attempts. When it does respond, it executes the command successfully. I’ve tried both “Alexa” and “ok Nabu” for the wake words, same behavior.
I just received my ESP32-S3-BOX-3. Same behaviour. I can set it up via the web page and it also is recognized in HA. But the Home Assistant Logo is greyed out on the device and for me no voice command is recognized.
What I noticed, is my breadboard esp32 (when connected) starts wake word runs that I can see in the Debug Assistant.
The Box3 does not (when selecting Home Assistant as the wake word engine location).
The Debug Assistant just stays empty. On any pipeline I try.
Not sure if it’s the same for other people, but for me wake word detection doesn’t work after rebooting the ESP32 until I switch “Wake word engine location” to “In home assistant” and then back to “on device”.
Might be some strange feedback, but hey. It’s progress.
Looking at the logs. I can see the last line is Waiting for VAD, or WakeWord. Depending on the Engine Location I set.
So… many frustrations later of this thing never detecting me, I spoke to it loudly at point blank close range… and it worked.
It has extremely poor pickup. Anything further than 10cm and it is unresponsive.
I don’t have a soft voice either.
So now the next goal would be increasing sensitivity.
Do I have two faulty units, or has anyone else struggled with the same issue?
