Every word, on the wall, as it is said.
Live captions for chapel, lectures, ceremonies and services. One app on one computer in your booth feeds the LED wall, the livestream, and every phone in the room — from the audio you are already mixing.
No per-hour vendor. No booking three weeks ahead. On the on-device engines, no audio leaves the building.
The wall view, running. Captions roll up from the bottom; the newest line is the bright one. It is a browser source — drop it into ProPresenter, OBS or vMix as a layer over whatever is already on screen.
Captioning is either expensive, late, or in the wrong place.
A CART writer costs by the hour
And has to be booked weeks out. That works for a scheduled accommodation. It does not work for the twenty-eight chapels, lectures and recitals on the calendar between now and December.
Stream captions miss the room
YouTube's auto-captions help somebody watching at home. They do nothing for the person sitting in row twelve, which is where the people who need captions usually are.
The proper nouns are the whole problem
Generic speech recognition does not know your president's name, your buildings, your donors, or the book of the Bible being read. Those are exactly the words an audience needs correct.
The same transcript, everywhere it needs to be.
The wall
A transparent browser source you add once as a ProPresenter Prop, an OBS browser source or a vMix title. Font, size, position, alignment and how many lines show are all set from the URL.
It never shows a spinner, an error or a placeholder on a screen an audience can see. If something breaks, the bar goes empty — not red.
Every phone in the room
Put the QR page on screen before the event starts. Anyone who scans it gets the transcript on their own phone, at their own text size, with scrollback — so arriving late or glancing away costs nothing.
No app to install, no account, no sign-in. It is a web page.
The livestream
Captions push into YouTube Live as the event runs. Afterwards the relay
hands you a timed .srt of every committed line, so the archive
video gets real captions instead of auto-generated ones.
Mark the video start on the status page and the file lines up with the recording.
Three pieces, and you already own two of them.
Audio you already have
A feed off the mixer you are already running — the same send that goes to the stream or the hearing-loop. USB interface, a spare output, whatever the booth already has.
One app, one computer
Install on the Windows or Mac machine already in the booth. It lives in the notification area or the menu bar. First run opens a setup wizard — microphone, engine, event title — then a guided tour that points at every control.
Captions out
The wall, the phones and the stream all subscribe to the same relay. Hand out the addresses, or just put the QR on screen. Nothing else to configure per event.
Built for whoever is in the booth that night
Usually a student worker or a volunteer, not an AV engineer. So: a preflight check that reads the network, the audio and the config and returns plain-English findings with one verdict — ready, ready with warnings, or not ready. A status page that shows whether audio is arriving and whether text is coming back. A rehearsal mode that replays a script with no microphone, no API key and no spend.
Names it will not get wrong
You give it a list — people, places, programs, the words peculiar to your campus — and the recognizer is primed with them before anyone speaks. There is a blocklist too, for the words you never want appearing eight feet tall behind a speaker.
You pick the engine — including ones that cost nothing and never touch the internet.
Speech recognition is a swappable part, not the product. Three real options ship, and the app substitutes a working one rather than failing if a config names an engine that machine cannot run.
| Engine | Cost | Runs on | Trade-off |
|---|---|---|---|
| Apple on-device | Free | macOS 26+ | Audio never leaves the machine. No network at all after the first model download. |
| Whisper on-device | Free | Windows, macOS, Linux | Also fully offline. Roughly a second more latency and less steady partial text. |
| Deepgram | Paid API | Anywhere with internet | Lowest latency and the steadiest partials. Your own account, your own key, billed to you. |
| Rehearsal | Free | Anywhere | Replays a script you write. No audio hardware, no key, no spend — for training a volunteer at 2pm on a Tuesday. |
More than one language
Optional. Offer a set of languages and every completed line is translated into each one. A phone that joins halfway through and picks Spanish replays the Spanish history, not the English. Only languages with a working translator behind them are ever shown — an audience never gets a dead choice.
Where the words go
The transcript lives on your machine and on your relay. The viewer pages are served from the same box. The public pages tell search engines to index nothing, because a live transcript of a room should never turn up in a search result.
It is in beta, and we would rather say so.
Automatic captions are not a legal accommodation. Where a student or guest has a documented accommodation requiring verbatim captioning, that is still a human CART writer's job, and your disability services office makes that call — not us. Campus Captions is for the much larger set of events that today have no captions at all.
Working today
All three stages are built and running real events — capture, relay, wall, phone, QR, status console, YouTube push, transcript export, translation, signed-off preflight. Installers exist for Windows and macOS.
Rough edges
The installers are not yet code-signed, so each shows a one-time warning the first time it runs. Accuracy depends on your room, your mics and the engine you choose. Reaching viewers outside your building needs a tunnel you set up.
No price, no pitch
It is not for sale. There is no pricing page because there is no price yet, and inventing one before the beta tells us what this is worth would be guessing at your expense.
What we want from a beta site
A real room with real events on a real calendar — and somebody willing to tell us plainly when it got something wrong. In return you get the software, the setup help, and a direct line to the person building it.
Tell us about your room.
We are adding beta sites slowly, so that each one gets actual attention. Leave your details and we will be in touch about whether it is a fit.
You're on the list.
We'll be in touch. If your events are coming up soon, say so in a reply when you hear from us and we'll move you up.
Before you ask
Does it work with ProPresenter?
Two ways. Add the wall view as a web Prop and lay it over your slides — that is the usual setup. Or send the captions into ProPresenter directly, with no web layer at all, if your operator would rather drive it from inside ProPresenter.
Do we need internet?
Not for the captions themselves, if you use one of the on-device engines. Audio never leaves the computer and nothing is billed.
You need the local network so the wall and phones can reach the relay, and you need internet only for the optional extras: the Deepgram engine, translation, pushing captions to YouTube Live, or letting people outside the building connect.
What hardware do we need to buy?
Most likely nothing. One Windows or Mac computer — the one in the booth is usually fine — and an audio feed from the mixer you already run. If there is no spare output, a cheap USB audio interface covers it.
How accurate is it?
It depends on three things you control: the engine you pick, how clean the feed off the board is, and whether you gave it your keyterm list. A clean lectern mic plus a list of your proper nouns is a different experience from a room mic and no list.
The honest framing: it is good enough that an audience follows along, and not good enough to be the official record of what was said.
Does our audio get sent somewhere?
Only if you choose the engine that does. Apple on-device and Whisper both run entirely on your machine. Deepgram is a cloud service and audio goes to them under your own account and their terms — that is the trade you are making for the lowest latency.
Can an audience read along in another language?
Yes, if you turn it on. Each completed line gets translated into the languages you offer, and each person picks theirs on their own phone. The wall stays in the source language.
Who runs it during an event?
Whoever is already in the booth. There is a status page that shows at a glance whether audio is arriving and text is coming back, a preflight check to run before doors, and a tour that walks a new volunteer through every control. Most of the job is starting it and watching one green dot.
What does the beta cost?
Nothing, and there is no price set for afterwards. If you run the Deepgram engine you pay Deepgram directly for your own usage; the free on-device engines avoid even that.