ESP32 S3 always requires wake word twice before activating/listening

[glad sinew]

Hi all, so I've been struggling to get some ESP32 S3's up and running, but I'm finally getting somewhere now. The only thing I'm missing is a minor, annoying problem. I have to say the wake word 'hey jarvis' twice each time I want to activate the satellite (and it's always twice. I've tried resetting, turning mute on and off, unplugging and plugging in...).

I can see that the ESPhome logs do show that the wakeword is being heard, but the assist in progress does not change to 'listening' until the second wake word. My config's attached.

Any help would be greatly appreciated 🙂

This is what I can see in the logs at the point of the wake word being heard (I can see that it's stopping the pipeline when hearing the wake word, but I'm not sure what I need to change in my config).

[21:06:39][D][micro_wake_word:160]: The 'Hey Jarvis' model sliding average probability is 0.980 and most recent probability is 0.988
[21:06:39][D][micro_wake_word:123]: Wake Word 'Hey Jarvis' Detected
[21:06:39][D][micro_wake_word:194]: State changed from DETECTING_WAKE_WORD to STOP_MICROPHONE
[21:06:39][D][micro_wake_word:129]: Stopping Microphone
[21:06:39][D][micro_wake_word:194]: State changed from STOP_MICROPHONE to STOPPING_MICROPHONE
[21:06:39][D][micro_wake_word:194]: State changed from STOPPING_MICROPHONE to IDLE
[21:06:39][D][voice_assistant:505]: State changed from IDLE to START_MICROPHONE
[21:06:39][D][voice_assistant:512]: Desired state set to START_PIPELINE
[21:06:39][D][voice_assistant:223]: Starting Microphone
[21:06:39][D][voice_assistant:505]: State changed from START_MICROPHONE to STARTING_MICROPHONE
[21:06:39][D][voice_assistant:505]: State changed from STARTING_MICROPHONE to START_PIPELINE
[21:06:39][D][voice_assistant:277]: Requesting start...
[21:06:39][D][voice_assistant:505]: State changed from START_PIPELINE to STARTING_PIPELINE
[21:06:39][D][voice_assistant:527]: Client started, streaming microphone
[21:06:39][D][voice_assistant:505]: State changed from STARTING_PIPELINE to STREAMING_MICROPHONE
[21:06:39][D][voice_assistant:512]: Desired state set to STREAMING_MICROPHONE
[21:06:39][D][voice_assistant:642]: Event Type: 1
[21:06:39][D][voice_assistant:645]: Assist Pipeline running
[21:06:39][D][voice_assistant:642]: Event Type: 9

[tawdry gazelle]

Can you try use_wakeword: false on VA maybe? Looks like you have both microwakeword AND openwakeword running - one on ESP, another in HA pipeline... 🙂

[glad sinew]

Just to be clear, is this in the ESP config under the 'voice_assistant' section?

[tawdry gazelle]

Yes