ChurchAudioAi
Live translation of the sermon into the languages of your congregation: speech is recognised, translated and voiced while the speaker keeps talking. This covers everything a church sets up and does, from signing up to the recording afterwards.
How the service works
One service involves three roles. They are worth telling apart, because each has different settings.
- Administrator
- Sets up the church, pays, prepares voices, keeps the glossary, invites the team. Works in the dashboard between services.
- Host
- On the day, opens the console, picks the sound source, starts and ends the translation. Needs a computer with a browser, not a phone.
- Listener
- Opens a link or QR code on a phone, picks a language, puts on headphones. Sets up nothing and registers for nothing.
Sound from the service reaches the system, speech is split into phrases at pauses, and each phrase is recognised, translated and voiced. The listener sees subtitles first and hears the voice after. The usual delay is six to nine seconds from the end of a phrase.
Understand this first
Translation runs only into the languages someone is actually listening to. You may offer all fifteen in a service — you pay for the ones people chose. With no listeners at all, nothing is generated and nothing is spent.
01Signing up and paying
/signup → confirm email → /onboarding → /dashboard
Signing up creates your church, makes you its owner, and credits $5 straight away — roughly 50 minutes of translation into one language. You can spend it immediately; no card is needed for that.
| What | How much | What exactly for |
|---|---|---|
| Welcome balance | $5.00 | Credited automatically at signup, once |
| Translation | $0.10 / min | For every minute into every language actually listened to |
| Minimum to start | $0.50 | Below this a service will not start |
| Church Pro | $99 / mo | Platform subscription, separate from the translation balance |
| Church Plus | $199 / mo | More simultaneous services, members and voices |
A common misunderstanding
The subscription and the translation balance are separate wallets. The subscription unlocks platform features; the balance pays for voiced minutes. An empty balance stops a service even when the subscription is paid.
Divide the balance by 0.10 for the number of minutes into one language. Two languages listened to at once halves it. The dashboard and the console show this estimate beside the balance.
02Dashboard
/dashboard
The main screen between services. Services are created here, past ones are listed, and the links to settings and billing live here.
- Create service — starts the creation wizard.
- Service list — past and planned, with links to the console, statistics and materials.
- Balance and subscription — top-ups, plan changes, invoices and card.
- Settings — organization, voices, quality, team.
The walkthrough
On a first visit a step-by-step guide starts: it highlights the control and explains what it does. Seventeen steps, from Create service to Stop service. Steps that ask for a press wait for that press rather than a Next button. Turn it off with the checkbox, bring it back from the console.
03Church profile and glossary
/settings/organization
This sets what shapes the meaning of the translation rather than how it sounds. Configured once, it applies to every service.
- Congregation name
- How you are named in materials and on the listener page.
- Denomination and tradition
- Loads the built-in terminology of your tradition, so settled expressions are translated the accepted way rather than literally.
- Service style
- Sermon, liturgy, prayer, Bible study, testimony, announcements, conference. Sets the register of the translation.
- Tone
- Neutral or informal.
- Spoken language
- The language the preacher usually speaks.
The church glossary is a list of terms with notes: ministers' names, department names, local turns of phrase, settled renderings of scripture. Each term can carry a note on exactly how to translate it.
Why this matters
Without a glossary a proper name or an in-house term easily turns into nonsense. Five to ten terms added in advance noticeably raise the quality of the very first service.
04Speaker voices
/settings/voices
The voice is what the translation sounds like in the listener's headphones. Catalog voices are ready immediately: one male and one female per language, with a male voice matched to the listener's language by default. Voices you record always win over the catalog and are visible only to your organization.
- Press New profile, give the speaker's name and a description.
- Add a sample: record in the browser with Record, or upload a file.
- Press Create and prepare, then wait for it to be ready.
- Press Test voice and listen to the sample phrase.
What the sample must be
Sixty to ninety seconds of ordinary, calm speech in a quiet room. No music, no echo, no second voice behind it. Singing and impassioned preaching make poor samples — the voice comes out uneven.
A profile you no longer need can be archived, and one that came out wrong can be prepared again from the same sample.
05Recording and quality
/settings/quality
The system can keep the original sound of a service and then check for itself how fully it was translated.
- Service recording
- On by default for new churches. It records the input — the same audio recognition heard.
- Automatic deletion
- After 7 days by default; adjustable from 1 to 90.
- Quality analysis
- Compares the recording with what was translated and reports live-text coverage and word difference.
- Issues found
- Places where the translation may have lost meaning, flagged for review.
- Download audio
- Take the original recording of the service as a file.
Consent is required
Recording can only be enabled after consent is confirmed — a requirement for handling a congregation's voices. Enabling it during a service starts recording from that moment: what has already been preached is not kept.
06Team and access
/settings/team
An invitation goes to an email address and the role is chosen at once. The invitation link can be copied and handed over in person.
| Role | What they can do |
|---|---|
| Admin | Everything: settings, billing, voices, team, running services |
| Operator | Create and run services, but not change billing or the team |
| Viewer | Only view services and statistics |
English only for now
The team page is not translated: its labels are in English — Invite a team member, Email, Role. Nothing breaks, but operators are worth warning.
07Creating a service
/create-event
- Name and speaker. The name is for you, to tell services apart. The speaker's name is what listeners see.
- Spoken language. The language the preacher speaks — not the translation language.
- Translation languages. What to offer listeners. Offer as many as might be needed: you pay only for the ones chosen.
- Continue. The service is created and the links appear.
- Listener link
- Public. This and the QR code go to the congregation.
- QR code
- The same thing, for a projector or a notice.
- Host link
- Opens the console and carries an access key.
Never mix up the links
The private host link hands over control of the service. It must not go to listeners — only the public one does.
08Console: starting a service
/host/<service id>
Open the console on a computer in Chrome. Work down the screen in order.
- Sound source: a microphone, or a browser tab carrying the stream.
- Translation mode: shows what is selected. Live translation is the default, and what services run on.
- Balance: the amount and how many minutes it buys. Below $0.50 the start is blocked.
- Check system: verifies balance, sound, connection, languages, voice, recording and providers.
- Start: the service begins, recording starts on its own, and the listener link goes live.
Mistake number one
When picking a tab, you must switch on Also share tab audio in the picker. Without it recognition receives nothing: the service looks like it is running and translates not a word.
09Advanced settings
A collapsed panel at the bottom of the console. Inside are three things you can hear on air: the quick mode, the recognition route and the speaking voice.
| Quick mode | When to choose it |
|---|---|
| Live translation | Ordinary services. Recommended and set by default |
| Highest quality | When coherent text matters more than speed |
| Conference | Several speakers, presentations |
| Sermon | Long, connected reasoning |
| Streaming | Sound from a tab, a stream |
- Nova-3 — the fastest recognition route by measurement, and the default.
- Flux — assembles larger meaning blocks and translates more coherently, but holds a phrase for 8–19 seconds. That is noticeable live.
- Open — the fallback when the main provider is unavailable.
In the voice picker your church's own voices come first, the catalog below. There is no separate language to choose: the voice follows whichever language the listener picked.
Fragment length
Longer fragments preserve meaning but produce more audio to speak. If listeners start falling behind and the gap grows, shorten the fragment. It is the only setting that changes this.
10What the listener sees
/listen/<service id>
- Opens the link or scans the QR code on a phone.
- Picks a language from those offered.
- Allows playback — on a phone the browser may ask for one tap.
- Hears the translation and sees subtitles; subtitles arrive before the sound.
- Restore translation
- When sound has stopped or the browser blocked it.
- Jump to live
- Skip the accumulated backlog and return to real time.
- Queue and lag
- Show how far behind the live speech this person is.
A phone in a pocket
Minimise the tab or lock the screen and playback stops while the queue keeps filling — the person starts falling behind. Ask the congregation to keep the tab open. If they have already fallen behind, Jump to live returns them to real time instantly.
Someone joining mid-service starts at the current point, not from the beginning. Headphones are required: sound from a speaker returns into the microphone and breaks recognition.
11During the service
The console shows figures that tell you whether everything is well.
| Reading | Normal | What it tells you |
|---|---|---|
| Text latency | 6–9 s | From the end of a phrase to subtitles |
| Audio latency | 7–10 s | To the start of the voice |
| Queue | 0–3 | Phrases waiting to be processed |
| Input level | moving | A meter standing still means no sound is arriving |
| Listeners | — | How many are connected, and in which languages |
Pause halts processing temporarily and Resume continues. At the end always press Stop service: that closes the session and saves the recording.
Do not just close the tab
Closing the console without stopping the service leaves the session open and the recording unfinalized, and the church ends up with no file of the sermon. The browser warns you on close — do not dismiss that dialog.
12After the service
/events/<service>/stats and /events/<service>/content
Statistics show how many listened, how many phrases were translated and what the average latency was. The content studio turns the sermon that was preached into finished materials.
- Structured outline — the points and shape of the sermon.
- Website article — connected text ready to publish.
- Social pack — short excerpts and quotations.
- Small-group material — questions for discussion.
The recording sits in the quality settings, can be downloaded, and is deleted automatically on the retention you chose. If a service was never stopped explicitly, its recording is marked partial: the audio in it is complete, the mark only means the system closed it rather than a person.
13Plans and limits
| Capability | Trial | Church Pro | Church Plus |
|---|---|---|---|
| Price | — | $99 / mo | $199 / mo |
| Languages at once | 15 | 15 | 15 |
| Services at once | 3 | 3 | 5 |
| Team members | 3 | 10 | 25 |
| Voice profiles | 3 | 5 | 15 |
A plan applies once it is paid
Until the subscription is active, trial limits apply — even when a plan is already selected on the card. Choosing a plan and paying for it are separate acts.
Translation minutes are not capped by plan: they are paid from the balance at $0.10 per minute per language, whichever plan you are on.
14What to do when
| Symptom | Cause | What to do |
|---|---|---|
| Service running, no translations | A tab shared without its audio | Re-pick the source and switch on tab audio |
| Start button will not press | A check failed, or balance under $0.50 | See which check line is red; top up the balance |
| A listener hears nothing | The browser blocked playback | Press Restore translation on the listener page |
| One listener far behind | Their tab was minimised or the screen locked | Press Jump to live |
| Everyone behind, gap growing | Fragments too long for the pace of speech | Shorten the fragment in advanced settings |
| Latency jumped past 15 seconds | First phrases after a pause or a start | Wait: connections are established on the first phrases |
| A listener cannot connect | The service is not running, or the wrong link | Check the service is live and hand out the public link |
| Names and terms come out oddly | The church glossary is empty | Add the terms in organization settings |