Live translation for church services: members scan one QR code and read or hear the sermon in their own language as it is preached. Setup, delay and cost.
Live translation for church services works like this: the sermon is transcribed once from the church's own sound, translated as it is preached, and each member opens one link on their phone, picks their language, and reads along or listens with earbuds. Nobody installs an app or signs up, and the service is preached exactly as before.
The rest of this article is about the people in the pews: how they join, what they hear, how far behind they are, and what a Sunday costs. The setup on the church's side (taking audio from the desk or the stream, adding the sermon notes, labelling speakers) is covered in Live transcription for churches, lecture halls and council chambers, so it is only touched on here.
The church creates one broadcast: a link and a QR code that stay the same every week. On Sunday someone presses Start on a live transcription attached to that broadcast. The audio comes from wherever the clean sound already is: the mixing desk, the livestream, or a microphone.
From then on, every sentence the preacher finishes is translated into the languages the people in the room have asked for. A member from Brazil reads Portuguese, the family from Kyiv reads Ukrainian, the student from Seoul reads Korean, all from the same transcript at the same moment. The preacher keeps speaking their own language and does not have to slow down or pause for anyone.
Because there is one transcript, everyone reads the same words, and a name corrected at the desk is corrected for everyone. A translation app on each phone hears the room from a different seat and makes its own mistakes. Why a phone's built-in translation is not enough for a room goes into the reasons.
No. A listener scans the QR code with the phone's camera, or types the short link, and the broadcast opens in the browser. There is no app to install, no account to create and nothing to pay. Visitors who walked in for the first time that morning can follow as easily as members.
Print the QR code in the bulletin, on a card in the pew racks and on a slide before the service. Because the link never changes, the same code works every Sunday, and if the broadcast is ever renamed, the old link still opens it. Before the host presses Start, the page says the broadcast has not started yet; once it goes live, the first lines appear on their own.
Each listener picks their own language from the 100 on offer. A single service can carry up to 10 languages besides the one it is preached in, which covers most congregations; the 10 are filled by the languages people actually choose, so nobody has to set them up in advance. A listener can switch language at any time and the sermon so far reappears in the new one. A Show original language setting puts the preacher's own words beside the translation, which helps members who are still learning the church's language and want to follow both.
Yes. Next to Read, the listener page has Listen: each translated sentence is read aloud, so a member can put in earbuds, put the phone away and hear the sermon in their language. That suits older members who find small text hard, and anyone who would rather look at the front than at a screen.
Whose voice it is depends on what the preacher has agreed to.
A cloned voice only ever reads its owner's own words, translated. The preacher's voice never reads the lines of the person who led the prayers, the scripture reader keeps their own voice, and no voice can be made to say something its owner did not say. Being a public figure on Sunday morning is not the same as consent, so ask before cloning anyone.
Voices cost nothing extra. One practical note: the first spoken lines of a service can take a moment longer than the text to start, and a few languages have text but no voice yet, in which case that listener reads instead.
A few seconds. The original-language text appears almost as the preacher speaks, the translation follows a second or two later, and the spoken version comes after its translation. Translation waits for the end of each sentence on purpose: languages order their words differently, and the end of a sentence can change how the beginning should read. A human interpreter waits for the verb too.
If the audio comes from a public livestream rather than the desk, the stream adds its own delay before VocalStack hears a word, sometimes several seconds. For the people in the room, a feed from the desk keeps them closest to the pulpit. Live translation latency, explained breaks down where each second goes.
The words a congregation notices most are names: books of the Bible, places, the guest speaker, the church's own ministries. VocalStack knows scripture names such as Habakkuk and Nebuchadnezzar. Words it cannot know in advance, like a visiting pastor's surname or the name of your youth group, are what the script is for.
Add the sermon outline, the readings or the slides as a script before the service, and the names and terms in it are written exactly as you spelled them. That carries through every translation, because every language starts from the same text. A volunteer can also correct a word from the host screen while the service is running; every listener sees the fix in their language.
Two honest limits:
Many churches already have a volunteer interpreting into the largest second language, either on stage beside the preacher or through headsets. Live translation does not have to replace them.
A common split is to keep the human interpreter for the language most of the visitors share and let the broadcast cover everyone else: the two families who speak Farsi, the Swahili-speaking student, the grandparents visiting from Poland. Nobody has to find a volunteer for each of those languages.
If the sermon is already interpreted on stage, sentence by sentence, the broadcast copes with that too. The spoken language is detected as it changes, and with speaker labels set to When Needed, the preacher and the interpreter each get their own label. If listeners turn on Listen, each is read in their own voice, never in each other's.
An interpreter knows the congregation and can ask the preacher what they meant. Software brings the languages nobody on the rota speaks.
Every service is kept. The broadcast page lists past recordings, so a member who was ill, travelling or on the nursery rota can open last Sunday's sermon, pick their language and read it or hear it read aloud. Once the service ends, a title and summary are written in every language, which makes a good line for the weekly email to members who read a different language at home.
The listener page also offers the transcript as a download, and the audio in their own language where a voice exists, for the drive to work on Tuesday. The church's library keeps every recording with its corrections and translations, searchable in any language.
You pay for the minutes of the service that are transcribed. Listeners, translations and voices add nothing, however many people follow and however many languages they pick. At launch prices, live transcription is $0.60 per hour of audio, billed by the second, with a 5-minute minimum per transcription on Pay as you go and none on Premium. The current rates are on the pricing page.
At those launch prices:
A new account comes with one-time signup credit and asks for no card, enough to try several Sundays before deciding. After that, a church tops up as it goes on Pay as you go, or subscribes to Premium, which includes credit every month. Two details decide which fits. Pay as you go has a maximum length per transcription, listed on the pricing page, so a long service or an all-day conference may need Premium. And a transcription on Pay as you go can take up to 15 seconds to start from cold, so press Start before the opening song.
It is not a human interpreter. It translates what was said, sentence by sentence, and it can be wrong: an unusual name with no script behind it, a preacher who drops their voice away from the microphone, two people speaking at once. Corrections from the desk help, but nobody should treat the text as a verified record of the sermon.
It is not sign language. Captions help many deaf and hard-of-hearing members, and they help people with hearing aids who struggle in a reverberant room, but a member whose first language is sign language may still need a sign language interpreter.
It is not a replacement for good audio. A phone on the lectern hears the room as much as the preacher, and the translation can only be as good as the words heard. Take the feed from the desk, as described in the setup article.
And it is public by default. Anyone with the link can read along, which suits a Sunday service. For a prayer meeting or a pastoral evening, make the broadcast private with a password, and say in the bulletin that the service is transcribed.
No. Listeners scan the QR code or open the link in their phone's browser. There is no app, no account and no charge for listeners.
At launch prices, $0.60 per hour of audio, billed by the second, so a 75-minute service is $0.75. Listeners, languages and voices are included. The pricing page has the current rates.
Up to 10 besides the language it is preached in, chosen from 100. Listeners pick their own; the first 10 picked are the ones the service carries.
Only if the preacher agrees. With their consent, their voice is cloned from about thirty seconds of their speech and reads only their own words, translated. Without it, a stock voice reads the translation.
Yes. A live transcription can take audio from a YouTube live, a stream address or an encoder, and churches that stream from OBS can use the OBS Studio plugin. For people in the room, a feed from the desk gives the shortest delay.
The spoken language is detected as it changes, so a bilingual sermon is transcribed in whichever language is being spoken and translated from there.
Create a broadcast at {{app}}/broadcasts/new, print its QR code, and run it at one service with a few members who read another language. The signup credit covers it. For a whole event with several stages, the multilingual conference checklist covers the rest, and One link, every language walks through the host and listener screens in detail.