Why a church translation tool fades, and what to compare next

If the live translation you tried last year is no longer running, the reason is almost never the translation itself. It is one of four things: the audio it was listening to, the names and verse references it got wrong, how much Sunday work it cost a volunteer, and whether anything was left of the service on Monday. Those four are what to compare before you choose again, and all four can be measured in your own room in one service.

We run Farsi and Armenian in Vienna every Sunday, so this list is not a theory. It is the order in which things actually break.

Why did the translation we tried get quietly dropped?

The feed comes first. If the tool was listening to the room mix or to the stream, it was hearing the preacher plus a band that is soft in one minute and full in the next, and the output wobbles with it. One dedicated feed carrying only speech fixes most of what people describe as bad translation. There are two ways to give it, and which audio feed live translation needs walks through both.

Then names. A congregation notices nothing faster than its own pastor's name coming out as a common noun, or Habakkuk arriving as something close but wrong. Every system slips there, because those words are rare in the language model and frequent in your building. What separates tools is whether a person on your team can approve a list of them and have it hold from week to week.

Third, the Sunday handling. If starting the translation takes a laptop, a login and someone who knows the trick, it survives as long as that person does. The question to ask is how many actions it takes between the opening song and the sermon, and who can do them when the usual volunteer is away.

Fourth, and this is the one churches rarely put on the list: what was left of the service afterwards. A tool that only translates is a cost that shows up every Sunday and nothing in between, and that is the kind of line a treasurer eventually asks about.

What should we actually compare?

Five things, in this order, and each has a number or a yes.

Delay. How far behind the speaker is the listener? Under a second feels like the person at the front is talking to you. Three or four seconds and people stop following a story. Hold a stopwatch on a sentence with a clear ending and write the number down.

Voice. Not whose voice, but whether it carries tone and rhythm. A translated service is still preaching, and a flat readout of correct words loses the half of a sermon that lives in the pauses. Have two listeners who speak the language rate it, because you cannot judge this in a language you do not speak.

Handling. Count the actions. Count them again with a different volunteer.

Glossary. Can a person on your team add your pastor's name, your building's name, the words your tradition uses, and have them approved rather than guessed? Ours is a glossary the team approves, and it is the single thing visitors comment on.

Cost per Sunday. Compare whole Sundays, not hours or minutes. Our numbers are on the pricing page and they do not change for being asked on a call.

How do we test it so the volunteers agree?

Use one real service, not a demo. Give it your own feed and your own room, with the doors open and the air conditioning on. Put two listeners per language in the seats with earphones, not in the office with headphones on a laptop. Write the ten names and references the preacher is going to use on a piece of paper before the service, then check them off afterwards. Ask the two listeners one question each: did you follow the story, and did it sound like him? If one of them speaks both languages, give them a third and narrower job and ask whether the instructions survived, because a story survives almost anything and an instruction does not. If you would rather carry the questions in one place than in a notebook, we keep six questions to ask any church translation software.

One practical thing we learned the hard way: the visitor who needs the translation most is the one who did not bring earphones. Keep a basket of cheap wired ones by the door, or hand them out at the welcome desk.

What has to still be there on Monday

This is where the second choice usually gets decided. Translation on the phone in the room is the part people see, and it is a third of what a Sunday can produce.

The same recording that carried the translation should give you, within the hour, a clean transcript, a summary, the full message, a small-group guide and drafts for your blog, newsletter and social, in every language you serve. A second QR code at the exit should open a screen where someone can leave a prayer request, say they are here for the first time, write a testimony, ask the sermon a question or read the message. Every one of those responses should land in one follow-up list with a status per person, so Monday is a list and not a memory. And the questions people ask should be answered from what was actually preached, with the passage cited.

A church with one language uses all of that every week, and switches the translation on the Sunday a second language walks in. That is why it stays switched on: the thing earns its place on the Sundays when nobody needs translating.

What to do this Sunday

Ask your desk operator for one aux send with only the speaking microphones on it. Write down the ten names. Put two people in the seats with earphones. Then compare the numbers you wrote with the ones you were promised.

If you want the same test run in your building with Farsi, Armenian, Ukrainian or anything else your congregation brings, we will set it up for one service and give you the recording and the numbers afterwards.