Audio
The controller that runs your machine also runs the room. Music in sync across the house, a doorbell that says who's there, a plant alarm that speaks itself in the right room and keeps asking until an operator acknowledges it — and LEDs that dance to the beat you're actually hearing.
Status Hardware-verified
Melanite includes a complete audio plane: synchronized playback across zones (the whole house in sync, or independent rooms), Spotify Connect and other sources, per-zone volume, mute and ducking, spoken announcements composed from live values and rendered in the listener's language, a push-to-talk intercom and paging, and real-time audio analysis — beat, BPM and a 16-band spectrum — for sound-reactive light. Speakers are ordinary devices: a blank ESP32 becomes a placed, zoned, playing speaker in under three minutes without touching a terminal. Alarms can route to speech, with repeat-until-acknowledged and quiet hours.
Walk between rooms and the music doesn't smear. Zones can play together or independently; the mixer is the clock, and a stalled source produces silence rather than drift.
A factory-fresh ESP32 becomes a placed, zoned, playing member speaker in under three minutes without a terminal — flashed, provisioned, adopted, positioned in the 3D scene, and chirping on demand so you know which box it is.
Compose a sentence once from live values and it renders two ways: on screen as text, and aloud with units, times and dates spoken as words — in German or English, per listener. The screen shows exactly what the room hears.
Route a diagnostic condition to speech by severity, scope and zone. It repeats until someone acknowledges it, defers during quiet hours, and stops queued speech the moment it's acknowledged — with who and when in the journal.
Push-to-talk between panels and browsers, and paging that ducks the music, plays a chime first, and reaches every zone speaker in sync.
Level, bass, mid, treble, a latched beat, BPM and a 16-band spectrum as live values — time-aligned so the LEDs match what your ear hears, not what the buffer holds.
Spotify Connect (your phone stays the remote), network stream sources, your own uploaded sounds, and two dozen built-in tones — chime, notify, confirm, warning, error, doorbell, siren — synthesized rather than shipped as a folder of samples.
Zone volume, mute, source and now-playing as HMI bindings, node ports, REST and assistant tools — all fed by a change-driven push channel, with no polling anywhere.
A plant fault at 02:00 must reach whoever is actually in the building.
How
The condition routes to speech: it announces in the workshop zone, repeats every twenty seconds until acknowledged, and is deferred out of the bedroom zone by quiet hours. An operator acknowledges it from the panel, a phone or your own HMI page — the speech stops immediately and the journal records who and when.
The bell should pause the music, chime, and announce the visitor in every room.
How
The bell input triggers an announcement built from a live value. Music ducks, a chime plays first, the sentence is spoken in sync across every zone in the listener's language, and the music comes back up. The same sentence appears on the wall panel.
Sonos gives you sync but knows nothing about your machine. A PLC can trigger a horn but cannot say a sentence. Melanite has both halves in one device: the audio plane is wired to the same live values as the control logic, so an announcement can contain the actual temperature, an alarm can insist on being acknowledged, and the LEDs can follow the beat the room is hearing.
Any speaker you can attach to a small Wi-Fi board — Melanite flashes and adopts ESP32-based endpoints in a guided flow — plus any audio output on the Melanite device itself (HDMI, USB or an I2S HAT). All of them play in sync as members of a zone.
No, except for internet sources themselves. Zones, announcements, the intercom and sound-reactive analysis all run locally. Speech can be synthesized entirely on-device; a faster voice on your LAN or in the cloud is optional, and losing it costs no added latency.
No — the audio plane is a module that is off by default and adds nothing to the real-time control loop when enabled. A machine build never pays for it.
Yes. That is the point of the text composer: you write one sentence with placeholders, and it renders as display text and as speech, with units, clock times, dates and ordinals spoken as words in German or English.
One binary runs every tier — the licence only toggles entitlements, never the feature set.