There are four ways to be heard in every room, and they differ less in sound quality than in what happens when nobody is standing there to make it happen.
The question that separates them is simple: is this a thing you do, or a thing that should happen on its own? Calling everyone to dinner is a thing you do, and a phone works fine. A reminder at 7:15 every school morning, a call to prayer at a time that moves daily, or a message that must play whether or not anybody remembered — those are things that should happen on their own, and that is where a phone stops being the answer.
Cast from a phone or a smart speaker. Google Home and Alexa both have a broadcast feature, and it is genuinely good for "dinner is ready". It needs the internet, an account, a phone in your hand and a microphone in every room you want to reach.
An intercom kit. Wired or wireless door-station kits with room units. Excellent for talking, poor at scheduling, and you are buying hardware for each room.
A wired amplifier and ceiling speakers. What a school or a shop has: one amplifier, speaker cable, and a microphone or an audio input. Best sound, most work, and on its own it does nothing without a person at the microphone.
A small computer running scheduling software. A Raspberry Pi or a spare PC connected to speakers you already have, running software that plays and speaks on a schedule and takes an announcement typed into a browser. This is what Multilarm is.
The shortest working version takes about ten minutes and costs nothing, assuming you already own a PC or a Raspberry Pi and a speaker. Install the software on the machine that is connected to the speaker, open its dashboard from any browser on your network, type the announcement, and press play. Four steps, no account, no purchase:
Nothing in that list involves an account, a subscription or a microphone in every room.
Reaching several rooms is a wiring question before it is a software question, and there are three shapes that work. One machine can drive several rooms through several audio outputs; several small machines can be kept in step across rooms; or one amplifier can feed ceiling speakers, which removes the problem entirely because there is only one source. Most houses end up with a mixture:
You can mix them. A house that has a kitchen speaker, a hallway amplifier and a Pi in a garage workshop is an ordinary configuration, not an exotic one.
You still can. The Web Remote is a web page, so any phone on the home Wi-Fi opens it and speaks to the house without an app. Push-to-talk sends your live voice through the speakers if you would rather talk than type. The difference from a smart-speaker broadcast is that the phone is optional: the schedule keeps running when the phone is in a coat pocket in another building.
If you want to say "dinner is ready" twice a week and nothing else, install nothing. A smart speaker you already own does that, and this page should not talk you into a project.
Scheduling software earns its place at the moment somebody says "it needs to happen even if I forget", or "it needs to be in the workshop but not the bedroom", or "I need to know it actually played".
Do I need the internet for this?
No. Scheduling, speech and playback all run on the device. A Multilarm machine with no internet connection at all keeps announcing indefinitely. The internet is only needed for remote control from outside the house and for updates.
Does it need a smart speaker in every room?
No. It drives ordinary powered speakers, an ordinary amplifier, or several small computers — whichever you already have. Smart speakers are one option among several rather than a requirement.
Can I record my own voice instead of using the synthetic one?
Yes. Any MP3, WAV, FLAC or OGG file can be scheduled or played on demand, and push-to-talk carries your live voice. Typed speech exists because recording a file every time you want to say something new gets old quickly.
Is it free?
For personal, non-commercial use at home, yes — the whole thing, permanently, with no feature held back. A one-off perpetual licence applies to schools, mosques, businesses and other institutions.
Will it wake the house up if it goes wrong?
Volume is a per-event setting rather than a system-wide one, quiet hours are honoured, and the schedule shows the next few events before they fire so you can see what is coming. Testing at low volume is the first thing the documentation tells you to do, and it means it.