Hey. I have the feeling there might not be a proper solution for this already, but maybe someone here has an idea.
I have two esp32-s3-box3 devices in separate rooms. it happens quite frequently that the “wrong” one is activated when I say the wake word.
is there any good solution to this? maybe by tweaking the wake word parameters?
thanks
Same issue here. But I put 2 speaker in my livingroom room.
There needs to be a solution to detect the speaker closest to the voice (possibly the sound intensity) and only activate one speaker in case both receive the hotword.
My temporary solution is to put 2 speakers with 2 different hotwords. Lol
yeah, I have the feeling there is no simple solution for this problem. most probably it would require the devices to communicate with each other, or with home assistant, and decide which one to activate based on a Voice Activity Detection value or similar.
You could have the wake word turned off when the room is not occupied. Back on when it is.
unfortunately I don’t have occupation sensors running
Triggering the wrong box3 is tricky. As a work around, you could set different wake words for each, by setting the wake word detection to ‘on device’, and changing the esphome code for each, but thats less than ideal.
As for the wrong speaker saying the reply, I would suggest using Robs s3-box firmware which has the esphome code in it for the s3 to actually show up as a media player and therefore be the responding device, using its built in speaker.
The current dev code also has support for tons of other features like timers, and volume control (the default s3 ephome code is set to be very quiet)
yeah that seems to be the only solution as of now, thanks.
I wonder if it would be possible to add a voice signal strength parameter to the trigger for the assist pipeline core/homeassistant/components/assist_pipeline/websocket_api.py at 570725293ca455083677f14c170c60f32a609f7c · home-assistant/core · GitHub
and from this, the assist pipeline does/does not initiate a pipeline run for the same wakeword on the different devices.
I assume though this could lead to problems for the same wake word being used on different floors for example.
and to be honest, I only had a brief look at the pipeline code and don’t have a complete picture as of now.
as for the alternative firmware, I agree it is nice and I have been following the thread. but at the moment it is also just nice to get OTA updates, without having to run the ESPHome CLI every time 
I have the dame issue. Wake word is often triggered on both boxes and the farthest do not hear the command.