
Axel
Arena announcer who has never used an indoor voice
Voice category
Announcer voices carry a scene's scale. A stadium, a train station, a tournament — all of them are established by one voice speaking to a crowd rather than to a person. Cast the announcement properly and you can skip a paragraph of description.

Axel
Arena announcer who has never used an indoor voice

Duke
Trailer weight for introductions, three words per breath

Noor
Station announcement clarity, designed to be understood

Nova
Automated in-world PA that is not impressed by you
Every voice is available on every plan — the library is not a paywall.
A scene set in a stadium and a scene set in a broom cupboard can contain exactly the same dialogue. What separates them is the acoustics and, far more reliably, a voice speaking to everybody at once. One tannoy line at the top of a scene does more world-building than a paragraph of description, because the listener has heard that kind of voice in a real building and knows immediately what sort of space they are standing in.
That is why an announcement is worth casting properly instead of treating it as set dressing. The announcer is not a character in your story, but they are a character in the scene: they have a job, a script they have read a thousand times, and an attitude towards both. The bored transit announcer and the delighted tournament announcer are doing the same work with opposite energy, and an audience reads the difference in about two seconds.
The arena register is loud, elongated and theatrical. Vowels stretch, the name lands last, and the pace drops at the moment of the announcement itself. There is a heavier variant of this — the trailer voice, three words per breath, all of them important — which works for ceremonial introductions and for anything you want the audience to take too seriously.
The transit register is the opposite. It is clear, even and faintly apologetic, built to be understood over noise rather than to be enjoyed. Transit announcements are dull on purpose, which means the comedy or the menace has to come from the content. A perfectly calm voice explaining that the station is closing because of something it declines to name is far more unsettling than any amount of growl.
The third register is in-world automation: the ship computer, the facility warning, the automated queue. This is the flat, unbothered one, and it is the register that most often carries plot, because an automated voice can say something appalling without reacting to it. It is also the easiest to overuse, so give it the two lines that matter and no more.
Real announcements have a shape: an attention phrase, the information, then a repetition of the essential part. "Attention, passengers. The nine-forty service to Halloway is delayed. The nine-forty to Halloway, delayed." That repetition is what makes it read as a genuine announcement rather than as a line of dialogue aimed at the audience, and it costs you six words.
Write each announcement on its own line so it becomes its own performance, and write pauses as sentence breaks rather than as stage directions. A parenthetical such as (over the PA) changes nothing about the render: a line that is entirely bracketed is skipped, and anything else is simply spoken. The PA feeling comes from casting plus your editor, never from a note in the script.
Numbers deserve a pass of their own. Platform numbers, gate letters, times and flight codes are exactly where a renderer's guess and your intention part company, and there is no pronunciation setting to correct it with. Generate the line, listen, and if "9:40" comes out wrong, respell it as "nine forty" and use that spelling everywhere in the script.
The most common failure is length, and it is worth repeating because everybody does it once. An announcement that runs six lines stops being atmosphere and becomes an information delivery system. The audience starts listening to the announcement instead of to your characters, and you have handed your exposition to the one voice in the scene with no stake in the story.
The second failure is casting the announcer from the same band as your leads. A tannoy voice has to sound like it comes from somewhere else, and the simplest way to achieve that is to pick a voice that is nothing like anyone in the room — different pitch, different pace, completely different attitude. If two characters end up sharing a base voice because the project is crowded, make sure the announcer is not one of them.
The third is processing too early. Everything that makes a PA sound like a PA belongs to the mix, not to the render. Keep the source clean and you keep your options: the same take can be a pristine broadcast booth in one cut and a corroded speaker in a flooded corridor in the next.
Short declarative phrases with gaps between them. An announcer addresses a room, not a person, so the sentences are shorter than dialogue and the pauses longer. Put each announcement on its own line — a whole paragraph pasted as one line is performed as one continuous thought.
A public address system is a midrange device, and the deepest voice in the library is not automatically the one that survives it. Audition on the actual announcement rather than on a neutral sentence, and keep the one that stays intelligible when you turn the volume down.
Announcements are slower than conversation because the room is noisy and the listener is only half paying attention. Line-level speed is available on Plus and Pro in five steps between 0.75× and 1.25×; one step down is usually enough.
Do not try to get the tannoy sound out of the renderer. Generate a clean read, then add the band-pass filter, the room reverb and the little burst of static in your editor, where you can tune them against the rest of the mix.
Two lines of announcement under a scene is atmosphere. Six is an exposition dump delivered by the least interesting voice present. If the audience genuinely needs the information, give it to a character who wants them to have it.
Band-pass filtering and a touch of room reverb in your editor, plus a click or a chime before the first word. The render should be clean so you can tune the effect against the rest of the mix — and so you can change your mind later about what kind of building this is.
They are opposite jobs. Arena announcing is elongated and openly theatrical, so a loud sports register or a heavy trailer voice fits. Transit announcing is even, clear and faintly bored, because it exists to be understood rather than enjoyed. Casting a stadium voice for a platform announcement reads as parody, which is sometimes exactly what you want.
Very. A stadium voice applied to something trivial is one of the most reliable comic setups available, and it works because the register is so familiar that the audience supplies the context for free. The joke is in the mismatch, so keep the delivery completely sincere.
Not from the renderer — it performs written lines, and crowd noise is sound design. Take a crowd bed from a library or record one, then lay it under the announcement in your editor. Stem export on Pro gives you the announcement on its own track, which is what makes that mix easy.
Yes, if you want them to sound different from everyone else. Each character in a project is a cast member, and the free plan allows two per project, Basic five, Plus and Pro no limit. A project also uses at most ten distinct base voices before characters begin sharing one, so in a crowded scene it is worth asking whether the announcement earns a slot.
On any paid plan, yes — commercial use is included in the licence for the generated audio. The free plan is personal and non-commercial. Inventing a fictional airline or arena is fine; imitating a real broadcaster's identity or a recognisable individual announcer is not.
Paste a script, let CastDub assign a voice to every character, and export a finished drama.
Create for freeNo credit card. Free gives you a one-time pot of credits.