Allow routing the audio output to a media player

[rhuanbarreto]

Describe the feature

I'm very satisfied with the Home Assistant Voice PE! But now I want to have better sound on its output. Of course, the aux output helps but I want to make the sound output in my soundbar (Samsung HW-Q995C) which doesn't have an aux input. For example, Alexa can output to it through the Smartthings integration. So, I thought Home Assistant could route the voice output to this device or any other integrated media output device.

Example commands

No commands. It should be some sort of configuration where we can set this output and it would pull the strings inside the HA event system to trigger an event to the selected media output.

Use cases

This would overcome the biggest pitfall of Voice PE: the sound output. By hooking into the power of home assistant, any connected speaker could be the output of the Voice PE device.

Anything else?

If there are any ways to do this today, please share. Couldn't find anything correlated in the forums.

[Zirilan]

I have been waiting for this since the voice assistant became a thing. I got a bunch of M5 echo devices since i planned to put them all around my apartment. The sound in them were horrible though and would really like to output to my Sonos speakers i have in every room!

[QsgNpC8]

In the case of SONOS, I have done this and it works well. You would need to:

• Enabled and configured Music Assistant (MA) in Home Assistant (HA) to use at least ONE SONOS speaker.
• Created a friendly name for that device (I chose [Area] Music Assistant for my naming convention to make sure the MA SONOS device name is unique from the HA SONOS device name if you chose to install HA that integration).
• In the MA configuration for the SONOS speaker, make sure the speaker is enabled and chose to expose this player to Home Assistant.
• I disabled the "Pre-announce TTS announcements" feature as I found the chime irritating. YMMV.

The MA SONOS integration is beautiful because it handles music ducking and resume easily using the music_assistant.play_announcement feature.

voice_assistant_sonos:
  description:  
    Redirects Voice PE device announcements to SONOS speakers in 
    pre-defined areas where a sonos speaker is assigned to the PE.
    Note - the 100ms delay ensures music is started so that MA
    can have something to snapshot after the announcement is made.
  sequence:
    - variables:
        # Map PE friendly names to Sonos entities
        sonos_mapping:
          "office assistant": "media_player.office_music_assistant"
          "alexs assistant": "media_player.alexs_music_assistant"
          "noahs assistant": "media_player.noahs_music_assistant"
          "guest assistant": "media_player.guest_music_assistant"
          "master assistant": "media_player.master_music_assistant"
        # Convert device name to lowercase for matching
        calling_device: "{{ device_name | lower }}"
        target_sonos: "{{ sonos_mapping[calling_device] | default('media_player.office_music_assistant') }}"
    
    # Ensures MA playback starts so we have something to snapshot
    - delay:
        milliseconds: 100
        
    - action: music_assistant.play_announcement
      data:
        url: "{{ url }}"
      target:
        entity_id: "{{ target_sonos }}"
        
  alias: voice_assistant_sonos
  mode: parallel  # Changed from queued to parallel for multiple PEs
  max: 5  # Allow up to 5 simultaneous calls
  fields:
    url:
      description: Audio-URL of Assist-Reply
      example: https://…/tts.mp3
    device_name:
      description: Name of calling PE device (from friendly_name substitution)
      example: "Office Assistant"
  icon: mdi:play-pause

After taking control in ESPHome Builder and choosing the "Edit" function, you would place this code below the code found on the PE . Make sure the PE "Friendly Name" is the name of the HA PE entity you choose to use in the "Substitutions" at the top. The HA script below will use this friendly name to tie to the associated SONOS in the area where the PE is located:

To use this, you would change the "sonos_mapping:" section to list the name(s) of own MA SONOS devices (HA >> Devices >> Music Assistant) -- again, verify that the MA friendly names (found in HA here) match the speaker friendly name in the MA interface. If not, the do so now.

If only tying one MA SONOS speaker and PE together, then use queued and delete the "max: 5" logic. I use multiple PE-SONOS ties, and if you will as well, then keep the "mode: parallel" logic and change the "max: #" to the number of PE devices you list in the "variables" section.

voice_assistant_sonos:
  description:  
    Redirects Voice PE device announcements to SONOS speakers in 
    pre-defined areas where a sonos speaker is assigned to the PE.
    Note - the 100ms delay ensures music is started so that MA
    can have something to snapshot after the announcement is made.
  sequence:
    - variables:
        # Map PE friendly names to Sonos entities
        sonos_mapping:
          "media player1": "media_player.media player1"
          "media player2": "media_player.media player2"
          "media player3": "media_player.media player3"
          "media player4": "media_player.media player4"
          "media player5": "media_player.media player5"
        # Convert device name to lowercase for matching
        calling_device: "{{ device_name | lower }}"
        target_sonos: "{{ sonos_mapping[calling_device] | default('media_player.office_music_assistant') }}"
    
    # Ensures MA playback starts so we have something to snapshot
    - delay:
        milliseconds: 100
        
    - action: music_assistant.play_announcement
      data:
        url: "{{ url }}"
      target:
        entity_id: "{{ target_sonos }}"
        
  alias: voice_assistant_sonos
  mode: parallel  # Changed from queued to parallel for multiple PEs
  max: 5  # Allow up to 5 simultaneous calls
  fields:
    url:
      description: Audio-URL of Assist-Reply
      example: https://…/tts.mp3
    device_name:
      description: Name of calling PE device (from friendly_name substitution)
      example: "Office Assistant"
  icon: mdi:play-pause

Again, this works well for me. YMMV. Use at your own risk of course.

It's quick and easy to go back to stock firmware if this doesn't work out for you. With a USB cable attached to your computer, just reflash the stock firmware: https://support.nabucasa.com/hc/en-us/articles/25800241218717-Reinstalling-the-firmware

[QsgNpC8]

To add a little more context to your request, the PE streams both the listening and replying steps into the pipeline. This allows for the pipeline to begin STT (converting your speech to text for computation) and TTS (converting response text to speech for playback). Because of that, the pipeline is designed to reply directly to the device/entity which recorded the request to save time.

Without streaming, both the request and response would be saved as an audio file each way and increase latency. With streaming, actions can start while the files are written. My code above is simply a workaround to use the audio file after response creation. What I have found, on my system at least, is the delay is negligible. There also considerations for long replies (ie, tell me a story) where streaming allows the reply to start now, versus loading the whole story while the user waits before beginning the response.

Currently, the only way around this process is to capture the url of the response audio file, silence the response of the PE and play the audio file as a (in my case using) Music Assistant announcement on the desired SONOS speaker. You could also choose a normal (non-music assistant call) media_player.announce feature and create your own logic for music pause and/or ducking (muting). Either way, a long response will make you wait.

Unfortunately, there is no way of separating the response from the pipeline without completely rewriting the PE firmware. You could absolutely take 100% control of the PE and fully rewrite the firmware as a new Wyoming satellite. In this manner, you would create your own pipeline and dictate exactly how the input, processing and output stages happen.

That's more time than I am willing to invest. This is an easier solution. Like you, I think it would be nice to have an option to choose the output device rather than the reply automatically come back to the PE without a choice.

My code is an extension of the work found here: https://community.home-assistant.io/t/redirect-voice-pe-replies-to-sonos/926652 Snapjack does a good job of making a helper to allow him to choose "where" his response goes too. Just a little light reading. Unfortunately, both of these code blocks are just workarounds. A native solution would be nice.

[ethagnawl]

I would love this option, too.

I was recently given an Ikea Symfonisk speaker. I've got an automation working which allows me to use it in a similar way to how I've been using my Voice Assistant Preview device as a media player but I don't want to reinvent the wheel, use aliases, etc. to get it 1:1. I really just want to use it as a dumb sink for the VA's audio output.

For the sake of completeness, I asked about this on Reddit before finding this GH issue: https://old.reddit.com/r/homeassistant/comments/1t9j6ye/replace_voice_assistant_speaker/