Announcements: say something once and every device in the house says it, in your own voice, with no Home Assistant and no internet in the middle.
Announcements
Speak into one device and the others chime and play what you said. The devices find each other over mDNS, so nothing is configured beyond a house word — the same word typed on each device's setup page, which is what makes them one house. A device with no word set neither sends nor takes them.
- On the Show, the drawer has an Announce tab. It shows how many other devices are listening. While recording, the strip along the bottom says so; a tap on it sends what you said and a hold throws it away.
- From an automation, the
announce_houseaction speaks the words you give it, or opens the microphone when you give it none. - Quiet hours are respected: inside them an announcement is shown and not sounded. Alarms, timers and calls are not affected.
Recordings are levelled before they are sent, so an announcement arrives about as loud as the music it interrupts.
Also new
- Quiet hours can be set from the screen rather than by editing
state.json. - A follow-up can be given a tone of its own, so the device asking again sounds different from the device starting.
- Tones the device makes about itself are silenced during quiet hours.
- The setup page takes the house word, and the radio rows have Play and Stop buttons so a station can be tried before it is kept.
Fixed
- The announcements service panicked on its first peer browse and the supervisor left it failed, so the device never found another device for the rest of that boot (#13). It looked like a network fault.
- The advertiser was started once per call to Run, stacking another mDNS registration of the same device on every restart.
- The peer list was replaced by every browse, so a device that missed one three-second look dropped out until the next one two minutes later, and an announcement sent in the gap quietly did not reach it. Devices are remembered now and forgotten only after missing several looks running.