Text to speech

Any Home Assistant text-to-speech voice, such as Piper or Home Assistant Cloud, can speak on a panel. Home Assistant renders the phrase to an audio URL, and the panel downloads and plays it through its POST /play endpoint. Copy the three blocks under Setup, then call script.ha_paneld_say.

Voices rendered by Home Assistant are far better than the text-to-speech built into older panels.

Play an announcement with the voice assistant on

If the panel has a microphone and its voice assistant is turned on, you do not need any of the setup below: assist_satellite.announce speaks a message on the panel directly, with no REST commands, secrets or scripts to configure. The rest of this page is for panels without a microphone, or without the voice assistant enabled.

Note

/play accepts the bare URL as the request body (or {"url":"…"}), not a form field. The API reference describes the endpoint.

Setup

You need a text-to-speech engine (for example tts.piper) and a long-lived access token (your Home Assistant profile, then Security) in secrets.yaml. You also write your Home Assistant’s local base URL into the script below.

secrets.yaml:

ha_token: 'Bearer eyJ…' # Long-Lived Access Token, prefixed with "Bearer "

configuration.yaml:

rest_command:
  # Render a phrase with an HA TTS engine; returns {"url", "path"}.
  ha_paneld_tts_get_url:
    url: 'http://127.0.0.1:8123/api/tts_get_url'
    method: POST
    headers:
      authorization: !secret ha_token
      content-type: application/json
    payload: '{"engine_id": "{{ engine }}", "message": "{{ message }}"}'

  # Send a ready audio URL to a panel's /play (raw URL in the body).
  ha_paneld_play:
    url: '{{ base }}play'
    method: POST
    payload: '{{ media_url }}'

scripts.yaml:

ha_paneld_say:
  alias: 'Panel: speak (TTS)'
  fields:
    message:
      description: 'Phrase to speak'
      example: 'Dinner is ready'
    target:
      description: 'Any ha-paneld entity on the target panel'
      example: 'light.living_room_screen'
    engine:
      description: 'TTS engine entity_id'
      example: 'tts.piper'
  sequence:
    - service: rest_command.ha_paneld_tts_get_url
      data:
        engine: "{{ engine | default('tts.piper') }}"
        message: '{{ message }}'
      response_variable: tts
    - service: rest_command.ha_paneld_play
      data:
        # Panel address from its HA device — no IP to hardcode.
        base: "{{ device_attr(device_id(target), 'configuration_url') }}"
        # Rebuild on your LAN base so the panel can fetch it (tts_get_url returns the external URL).
        # Inline your internal_url here (templates can't read secrets.yaml):
        media_url: "http://homeassistant.local:8123{{ tts['content']['path'] }}"

Tip

Templates cannot read secrets.yaml, which is why the local base URL is written into the script rather than kept as a secret. Either write your internal base URL into the script as above, or expose it through a template sensor. If your external URL can be reached from the panel, you can skip the rebuild and use tts['content']['url'] from the response directly.

Use it

From an automation or the Home Assistant interface:

- service: script.ha_paneld_say
  data:
    message: 'The washing machine has finished'
    target: light.living_room_screen # any entity belonging to the target panel
    engine: tts.piper

The panel’s volume is number.<panel>_volume; set it beforehand if needed. To speak in several rooms, call the script once for each panel.

How it works

1. Home Assistant renders the phrase and returns an audio URL from POST /api/tts_get_url.
2. tts_get_url returns your external URL, so the script rebuilds it from your internal_url plus the returned path, which the panel can fetch on your local network.
3. The script posts that URL, raw, in the body of a request to the panel’s /play. The panel’s address comes from the configuration_url of its Home Assistant device (http://<ip>:8888/), so all you pass is an entity belonging to the panel.