Skip to main content
PIPELINE & ARCHITECTURE

Live translation, without the complexity.

If you can plug in a microphone, you can run Pentecost. Here is how the whole pipeline connects from speaker to audience.

1. Connect Audio

Audio comes in from your mixer (via a USB audio capture device), wireless mic (e.g. DJI Mic Mini), or straight from a Zoom/Teams meeting — no virtual audio cable or driver to install. Connect via our Windows / macOS app.

2. Pentecost Spark Cleans and Translates

Spark filters false starts and made-up output, then an AI cleaner fixes mishearings against your sermon manuscript and translates each line into every caption language at once, using your glossary and your Bible translation. Every line gets a confidence score, and doubtful lines are retried on a stronger engine.

Diagram of what Pentecost Spark does to every line: route table, health scores, speech to text, guards, domain pack, LLM cleaner, confidence routing, backup engine and caption out.

More about Pentecost Spark →

3. Instant Multilingual Delivery

Captions update instantly on attendee smartphones (QR scan), main auditorium projectors, OBS/ProPresenter stage overlays, and Zoom/Teams meetings.