Your service, translated live, in the ear of anyone who needs it — on the phone already in their pocket.

How it works

1. The church starts a session. Audio comes from a browser on a laptop at the sound desk. If you need particular channels from a Dante network or a multichannel mixer, it comes from the Global Pulpit Bridge app on the sound-desk computer instead.
2. Listeners scan a QR code. Put the code on a lobby sign, in the bulletin, or on the screen. It opens a web page. There is no app for the congregation to install. If their phone is set to a language you are offering, it is already chosen; otherwise they pick one. Then they put in their earbuds.
3. They hear the sermon in their language. Spoken aloud, with live captions running alongside. The person at the desk can drop chapter markers — “Announcements”, “The Message” — and they appear in every language at once.

Spanish

Portuguese

Vietnamese

Your first Sunday

There is no site visit, and unless you use the Bridge app there is nothing to install. Start to finish, it is an afternoon of setup and then a normal Sunday.

1. Create an account and add a card. You are invoiced once a month for what you used, and the card is how that invoice gets paid. Nothing is charged until you translate something.
2. Choose your languages. Turn on the ones your congregation might need. A language only runs, and is only billed, while at least one phone is listening to it, so offering one that nobody picks this week costs nothing that week. You can change the list whenever you like.
3. Print the lobby sign. Global Pulpit generates the QR code and the sign for you. Put it where visitors will see it on the way in, or drop the code in the bulletin.
4. On Sunday, go live. Open the Broadcast page on the laptop at the sound desk, choose your audio input, and press Go live. While it runs you can see how many people are listening in each language, and read each translation as it goes out. Press Stop when the service ends.

If you need particular channels from a Dante network or a multichannel mixer, see Global Pulpit Bridge below.

What it costs

There is no monthly fee. If you do not hold a service, you are not charged. We bill for one thing:

A translator-minute is one minute of one language. Three languages running for an hour, each with someone listening, is 180 translator-minutes.

You choose one of three voices in Settings, and it applies to every language. The standard voice is the fastest, about a second behind the speaker. It follows the preacher’s delivery, so the voice can shift during a service. The consistent voice (Settings calls it Hybrid) gives every listener one steady voice instead. It waits for each thought to finish before speaking, so it runs a few seconds further behind the speaker, and it runs a more expensive pipeline, so those minutes carry their own rate.

The economy voice also speaks in one fixed voice, or in a voice modeled on your preacher’s, and costs the least per minute. It runs furthest behind the speaker, often six seconds or more. It is offered only while it costs less than the standard voice.

Every rate is twice what it costs us. That 2× markup is the whole business model. We are not hiding it in a platform fee or a per-seat price. The margin covers running the service and keeping it supported; we keep it thin so the number stays small for churches.

Because our cost moves, your rate moves. The models underneath are priced by the companies that run them, and those prices change. When one comes down, the rate you pay comes down with it. When one goes up, we will tell you before your rate follows. We would rather say that plainly than promise a fixed price we cannot keep.

A language is billed only while someone is listening

Each language starts translating when the first person picks it, and stops about fifteen seconds after the last one leaves. If nobody chooses French this week, French costs nothing, even though you offer it. While someone is listening, the meter runs for as long as you are live, pauses and prayer included, so stop the session when the service ends.

A worked example

A 75-minute service with two languages, each with someone listening from start to finish, at the standard voice.

For reference, our own cost on that service is 150 × $0.034 = $5.10.

To estimate your own at the standard voice: minutes you are live × languages someone is listening to × $0.068.

Invoiced monthly for what was used. No minimum, no seat count, no contract, no setup fee.

Languages

Thirteen languages are supported today. Any number of them can run in a single service, and each one is billed only while someone is listening to it.

• English
• Spanish
• Portuguese
• French
• German
• Italian
• Indonesian
• Vietnamese
• Russian
• Chinese
• Japanese
• Korean
• Hindi

Privacy and recordings

Churches ask this before anything else, so here is the whole picture rather than a reassuring sentence.

Listening asks nothing of your congregation

Listeners never make an account, never sign in, and are never asked for a payment. The page counts unique listeners using a random identifier stored in their own browser. With it we record which language a device chose and how long it listened, so the church can see how many people listened, in which languages, and how many came back. It is not tied to a name, an email address or an account, and the church sees only totals, never individual devices.

There is one thing a listener can choose to give, and only by typing it in themselves: an email address, to be told when a service starts. It is the single exception to the paragraph above, and it is opt-in twice over — see below.

Being told when a service starts

A church can keep a list of people who want an email the moment a service goes live. Getting onto that list takes two deliberate acts by the person themselves: typing their address in, then clicking a link in the email that follows. Nothing is sent in between.

A church can invite someone, but it cannot subscribe them. An invitation is an email asking whether they want this. The church can ask again, but nobody is on the list until they click the link themselves.

Every notice carries an unsubscribe link, and unsubscribing takes effect immediately. After that the church cannot subscribe them again. If you change your mind, enter your address on the church’s listen page again and confirm it again, the same two steps as the first time. It is your decision to reverse, not theirs.

The list is the church’s and nobody else’s. We do not sell it, and we never write to it for our own purposes. The addresses pass through Postmark, the service that delivers these emails, and go nowhere else.

What is kept from a service

When a session runs, Global Pulpit keeps the translated audio for each language and a transcript of each translation. That is what makes a service available to replay afterwards. A language is only translated while someone is listening to it, so a language nobody picked leaves nothing behind. The original-language audio is never kept. On the standard and consistent voices, no transcript of it is kept either. The economy voice works by writing down what the preacher says before translating it, so on that voice a transcript in the original language is kept with the service, under the same retention window and the same delete button as everything else.

Only people signed in to your church’s account can list, play or download any of it. There is no public link and no shared address; the listening page carries the live service only, never the archive.

How long it is kept

You decide. An account owner can choose a retention window of 30 days, 90 days, six months, one year, two years or seven years, and anything past it is deleted automatically, audio and transcripts together. The default is to keep everything until you say otherwise, so nothing disappears unless you ask for it to.

An account owner can also delete any single service from the sessions list, whenever they like. That is immediate and permanent.

A preacher’s voice

On the economy voice, a church can upload up to two minutes of its preacher speaking, and listeners hear the translation in a voice modeled on theirs. Only an account owner can upload one, and only after confirming they have that person’s permission. We record who confirmed it and when.

The recording goes to Soniox, which runs the economy voice, to build the voice. Global Pulpit does not keep a copy; we keep only a reference to the voice Soniox made. Removing it in Settings deletes the voice at Soniox too. Services translated in that voice are kept like any other, under your retention window.

Where the audio goes

Producing a translation means sending your service audio to the company whose models do the work, and which company depends on the voice you choose. On the standard voice it is OpenAI. On the consistent voice OpenAI translates, and the translated text, not the audio, goes to Cartesia to be spoken. On the economy voice the audio goes to Soniox, which transcribes, translates and speaks it, and your Preacher’s words list goes with it. Session titles and chapter markers are short pieces of text, and OpenAI translates those on every voice. Under OpenAI’s terms for API customers, what it receives is not used to train its models. Soniox and Cartesia handle what they receive under their own terms for business customers.

The live audio travels between your church, your listeners and us through our own media server, not a third-party streaming service. The translated audio we keep is stored with Backblaze, and transcripts and account details are kept in our own database.

If your church needs this in writing, or needs something narrower than the above, ask. We would rather answer a hard question before you commit than after.

Global Pulpit Bridge

Most churches can start a session from a browser and a decent microphone input. A browser only hears the first two channels of an audio device, though, so it cannot pick the pulpit mic out of a Dante network or out of a mixer that sends all its channels to the computer. For that there is a companion app called Global Pulpit Bridge, for the Mac, with Windows and Linux in preview. It shows every channel with its own level meter, and it mixes the ones you unmute into the audio that is sent for translation: the pulpit mic on its own, say, or the pulpit and a lapel mic without the worship band. It is optional, and it does not change how you are billed.

The volunteer at the desk is often the one who steps away to greet someone or pray with them. While a service is live, the Broadcast page shows a QR code next to the channels. Anyone on your team can scan it, sign in with their own account, and mute or unmute channels from their phone, anywhere in the building, over Wi-Fi or cellular. Changes take effect right away, and the phone says so plainly if the broadcast has gone silent or it has lost touch with the desk. The phone only changes what is sent for translation, not what the room hears, and it cannot start or stop the service.

Muting every channel, from the desk or anywhere else, stops the translation in every language straight away instead of letting the last few seconds play out. That needs Bridge version 0.23 or later.

Stream Deck

If your desk has a Stream Deck, it can run parts of the service through Bitfocus Companion (version 4.3 or later). An account owner makes an access key for it under Settings, where the Global Pulpit module for Companion is linked too. Every key shows whether you are live and how many people are listening. The owner chooses what else it may do: mark chapters and set the service title, turn each language’s volume up or down, mute and unmute the Bridge’s channels, and go live and stop the service, which both take a second press to confirm. Going live from the deck needs the Bridge, with Companion running on the same computer. It starts the service just as Go live in the app does, and subscribers get their email; if a service is already live, the deck leaves it running.

Download the Bridge

• Mac with Apple Silicon (M1 or later), macOS 15 Sequoia or later
• Mac with an Intel processor, macOS 14 Sonoma or later
• Windows 10 or 11 on Intel or AMD (preview)
• Linux, Ubuntu 22.04 or later or Debian 12 or later (preview): Intel or AMD, or 64-bit Arm

Version 0.23.0. Not sure which Mac you have? Open the Apple menu, then About This Mac: a chip called M1 or later means Apple Silicon.

None of these are signed yet, so each system asks once before the first launch. On a Mac, macOS blocks it: open System Settings, then Privacy & Security, scroll down, and click Open Anyway. On Windows, SmartScreen says it protected your PC: click More info, then Run anyway. Windows 11 with Smart App Control turned on blocks unsigned apps entirely, with no way past it; write to hello@globalpulpit.com if that is you. On Linux, install it with sudo apt install ./global-pulpit-bridge_amd64.deb, or the arm64 file.

After that, open Global Pulpit in Chrome on the same computer and sign in. If Chrome asks to connect to devices on your local network, allow it, and the two connect by themselves. Safari cannot reach the Bridge. On a Dante network the computer also needs Dante Virtual Soundcard or a Dante interface, because the Bridge can only hear what the system itself can see.

Windows and Linux are new, and each has a limit the Mac does not. On Windows, Dante Virtual Soundcard has to be set to WDM mode, not ASIO, and then gives the Bridge two channels at a time, so in Dante Controller route the channel or the mix you want to its Receive 1–2. On Linux, a multichannel input that goes through the desktop sound server can lose channels, so use a stereo pair.

Questions

How much delay is there?

It depends on the voice you choose. The standard voice runs roughly a second or two behind the speaker. Listeners generally follow along without trouble; it feels like a slightly slow interpreter, not a recording. The consistent voice waits for each thought to finish before speaking, so it runs a few seconds further behind. The economy voice runs furthest behind, often six seconds or more. That delay is the trade for one steady voice, or for the lowest rate.

Does the congregation need to install anything?

No. The QR code opens a web page in their phone's browser. They choose a language and listen. Earbuds or headphones of their own are all that is needed.

What about the church's own equipment?

A computer at the sound desk with a network connection and a way to get service audio into it. For most churches that is the whole requirement. If you need particular channels from a Dante network or a mixer’s multichannel feed, you also need the Bridge app, which runs on a Mac (Apple Silicon or Intel), with Windows and Linux in preview. See Global Pulpit Bridge.

Can someone mute a mic for listeners without going back to the desk?

Yes, if the desk uses the Bridge app. While a service is live, anyone on your team can scan the QR code on the Broadcast page, sign in with their own account, and mute or unmute channels from their phone, over Wi-Fi or cellular. It changes what is sent for translation, not what the room hears through its speakers. It does not work for a service running from a browser microphone. See Global Pulpit Bridge.

Is this a machine translation, or a human interpreter?

It is machine translation. On the standard voice, a speech model listens and speaks in one step. The consistent and economy voices turn the speech into translated text first, then read it aloud in one steady voice. Either way it is very good and it is not perfect. It will occasionally mishandle a name, an idiom, or an unusual turn of phrase. We would rather you know that going in.

What if we see a translation going wrong?

While you are live, the Broadcast page shows each language’s translation as it goes out. If one is going wrong, mute every channel in the Bridge (from the desk, a phone or a Stream Deck), or every microphone on the Broadcast page. Every language stops for listeners straight away, without playing out the few seconds already on their way, and unmuting picks up with what is said next. Muting only some channels just takes them out of the mix. With the Bridge, this needs version 0.23 or later. What was translated before you muted stays in the recording, and the meter keeps running while you are muted.

Can we control how names and church terms come out?

On the consistent and economy voices, yes. You keep a short list for each language: a word the translation uses, and the word your congregation should hear instead. It swaps whole words only, so for a verb you add each form you use. On the economy voice you can also steer from the preacher’s side: when the preacher says a particular word, the translation strongly prefers the wording you set. That is a strong preference, not a guarantee. None of this works on the standard voice.

Can the translation sound like our preacher?

On the economy voice, it can. Upload up to two minutes of the preacher speaking, clearly and on their own, and listeners hear the translation in a voice modeled on theirs. You decide service by service whether to use it, and it costs the same as the economy voice. It needs the preacher’s permission, and the upload asks you to confirm you have it. See Privacy and recordings for what happens to the recording.

Who pays — the church or the listener?

The church. Listeners are never asked for anything.

What happens if we skip a month?

Nothing is invoiced. There is nothing to cancel and nothing to pause.

Can we cap what we spend?

There is no spending limit to set, but you control what the bill depends on. The voice you choose sets the rate per minute, from the economy voice at the low end to the consistent voice at the high end. Each language runs only while someone is listening to it, and only while you are live. A session left running by mistake stops itself after 90 minutes of silence, and no session runs longer than four hours.

Is the audio recorded?

The translations are. The translated audio for each language is kept, along with a transcript of it, so a service can be replayed afterwards. The original-language audio is never stored. A transcript of the original is kept only on the economy voice, which works from one. Only people signed in to your church’s account can reach any of it. You set how long it is kept, and an account owner can delete any service at any time. See Privacy and recordings.

Can we use it for something other than a Sunday service?

Yes. Anything with a speaker and an audience works the same way: a midweek study, a conference session, a funeral, an annual meeting. Billing does not change with the occasion. A single session runs for up to four hours, so a longer day needs a fresh start between sessions. A church has one service live at a time: pressing Go live on the Broadcast page of a second computer ends the one already live.

What if the internet drops mid-service?

A drop of a few seconds is usually ridden out. If the connection at the church stays down longer, the session ends. Once it is back, the Broadcast page says the service ended, and someone at the desk presses Go live again. Listeners do not have to scan again: if their stream drops, the page says it is reconnecting, in their own language, and keeps trying to rejoin for up to five minutes. The meter stops soon after your audio stops reaching us, so an outage costs you next to nothing.

How many people can listen at once?

There is no per-listener charge and no seat count. Whether eight people listen or two hundred, the bill is the same. It depends on minutes and languages, not on how many phones are connected.

Get in touch

Global Pulpit is new and we are working with our first churches now. You can set an account up yourself and run a service whenever you like; nothing needs to go through us first, and the Bridge app is a free download.

If you would rather talk it through, or want a walkthrough of a session before deciding anything, send an email. A person will answer it.

Create an account hello@globalpulpit.com