Method of disabling conversation response in an automation

Configuration Voice Assistant
1

Hi,

I’m in the process of setting up in house audio, with Home Assistant, Music Assistant and a ESP32-S3-BOX-3 and a Voice PE.

To implement this I am creating automations that use the conversation trigger, the automation then:

  • Determines which area of the house the Assist Satellite that answered the request is in with the area_id template function
  • Plays the requested music on a media player in the same location as the Assist Satellite.
  • With my current setup the Assist Satellite and the Media Player are the same device.

This works with no issues using a Voice PE, however I’m having some problems with the ESP32-S3-BOX3.

When the automation targets an ESP32-S3-BOX3 music plays for a second, however then the conversation response of “Done” is then sent which stops the music playback.

I’ve been trying to find a method of disabling the conversation response completely to avoid this issue. So far I’ve added different iterations of the Set Conversation Response action to things like "", "." and . like other forum posts have suggested, but these still result in a conversation response being sent which stops the music playing on the ESP32-S3-BOX3.

Has anyone found a way to completely disable a conversation response in an automation?

Thanks!

2

I don’t have a BOX3 to test with, but you might want to try having your automation call a script instead of calling the media play action directly. that would give you the opportunity to start the script with a short delay, but the automation would complete itself and run the conversation response.

3

Do you have the latest firmware version? The announcement sound shouldn’t interrupt the music if you’re using a modern dual-pipeline configuration.
Regarding the empty response in automation, you need to switch to YAML mode and set the two quotation marks there.
set_conversation_response: ""

4

@mchk

I’ve also set the conversation response in YAML as you’ve described to avoid the character escaping that is added if entered in the GUI, the same issue occurs, music stops and it sounds like a second of “static” is sent to the BOX3.

For the config, I’m on ESP Home 2025.11.5. I use the package at

github://esphome/wake-word-voice-assistants/esp32-s3-box-3/esp32-s3-box-3.yaml@main

And then add my additional config for the touchscreen in there so I believe everything is up to date. I’ve noticed in general with the BOX3 that if I manually start music playing on Music Assistant I can’t use it as an assist satellite until I stop the music playing, (it doesn’t respond to the wake word, and even the HA icon that is normally present on the screen disappears, once music is stopped the HA icon reappears and assist works as expected)

My work around for that was to activate the red action button on the BOX3 touch screen and connect it to an action that stops music.

@Didgeridrew
This is a great idea, I’m keeping this in the back pocket to use as a workaround if I can’t match functionality with the Voice PE, thanks for the suggestion!

5

This configuration version has one channel for announcement. Try searching for other firmware versions.

6

Thanks I’ll have a look, would you be able to share how you determined that config only has a single channel available?

7

This is defined in the speaker and media_player sections.

This is not the most up-to-date version of the VPE configuration, which uses an even more complex configuration. But it will suffice for illustrative purposes.

8

Thanks for this, I’ll do some digging and see what I can do.