Five things to settle before you pick AI translation

When a church types "AI translation for churches" into a search box, it is almost never asking a single question. Behind that one phrase sit five decisions, and most of the disappointment we hear about afterwards comes from a church that compared prices before it had made them. Settle these five first and the shortlist shrinks to one or two before you book a single call.

Who in the room cannot follow, and when did you last count?

Start with people, not with features. A church that counts gets a very different answer from a church that estimates. Walk the room one Sunday and ask three questions: how many people are following the sermon in a language that is not their first, how many of those are over sixty, and how many arrived in the last three months. The over-sixty answer decides whether you need a bigger QR code on the printed card and someone standing by to help. The newcomer answer decides whether you are solving for regulars who already cope or for visitors who will not come back if they understand nothing.

Our own Sunday in Vienna runs German, Farsi, Armenian and English. The number that changed how we built everything was not the total — it was that the people who most needed the translation were the ones least likely to ask for it.

Live in the room, or the sermon afterwards?

These are two different products and churches routinely buy one while needing the other. Live means the words reach the listener while the sentence is still warm — under a second behind the speaker, on their own phone, with captions alongside for anyone who would rather read. Afterwards means the recording comes back later, dubbed or written, in the languages you serve.

Plenty of churches need both, but they are paid for differently and tested differently, and a system that is excellent at one can be ordinary at the other. If you are not sure which you need, ask what happens on the Sunday itself: does the person sit there understanding nothing for forty minutes, or do they follow well enough and simply want the message again during the week? The first is a live problem. The second is a record problem, and it is cheaper.

Voice, text, or both?

A translated service can arrive as captions to read, as speech to listen to, or as both at once. Reading works for a short talk and fails for a long one — the eyes get tired, and a person reading a screen is not looking at the preacher. Speech works for a sermon and is the only thing that works for someone whose reading in any language is weak, which includes more newcomers than most churches assume.

Ask what the voice actually sounds like before you decide. A natural voice that keeps the preacher's tone and rhythm carries an exhortation; a flat text-to-speech read of a transcript does not, and your congregation will tell you so within three weeks. Most systems now offer both, so this is a quality question rather than a category question: what to test before you choose is the honest way through it.

What do listeners hold in their hands?

Three answers are in common use and they cost wildly different amounts. Receivers and headsets, bought and handed out: reliable, but somebody charges them, cleans them, chases the three that walked home in a coat pocket, and you need as many as your largest Sunday. Your interpreter, live in a booth or on a channel: still the best thing in the world for one language when you have the right person, and nothing in software replaces them. Or the listener's own phone, opened from a QR code with nothing to install.

The phone route is the one people underestimate in both directions. It is far less work than it sounds — no inventory, no charging, no sanitising — and it has one real limitation worth planning for: earphones. The visitor who needs the translation most is the one who did not bring any. Put a basket of cheap wired earphones by the door and the problem disappears for about the price of one receiver. That, plus the audio feed the system actually listens to, is most of what makes a phone-based service work.

How will you know it is any good?

This is the question almost nobody asks before buying and everybody asks three months later. Decide the test now, while you are still choosing.

The check that works costs one person and ten minutes. Ask somebody in your congregation who speaks both languages to listen on a phone during the sermon and mark only the imperatives — the sentences that tell people what to do. Not the whole sermon, not the stories. Narrative survives almost anything; it is the conditions, the negations and the exhortations where a translation can stay perfectly fluent and say the opposite of what was preached, and nobody in the room who speaks only one language will ever catch it. Have them check more than once, and above all during the part of the sermon that applies the message.

Then ask what the system does with names and terms. Every church has forty words that matter — the names of your people, your ministries, the way your tradition says a handful of theological terms — and a glossary that a person on your team approves is the only reliable way to keep them right week after week. If a vendor cannot show you where that list lives and who edits it, you have learned something.

Then, and only then, compare

Once those five are settled, a comparison takes an afternoon rather than a month, and the six questions to ask any provider do the rest. Ask for a trial on a real Sunday rather than a demo, and run the ten-minute check on it.

We built Sanctavo for our own four-language service before it was a product: one QR code for live translation on the listener's phone, the same code carrying Amen, a prayer request, first time here and a word to the team; a second code after the service that opens prayer, testimony and questions about the message, with every response landing in one follow-up list; and within the hour a transcript, a summary, the full message, a small-group guide and drafts for blog, newsletter and social in every language the church serves. Churches with one language use the after-service screen, the follow-up list and that content pack every week, and switch translation on the Sunday a second language walks in. Prices are public at our pricing page.

Whatever you choose, do the counting first. The five answers above are yours regardless of whose software ends up in the room.