How Gibberly works
By Gibberly Team · 15 January 2025
When a pastor steps up to the pulpit, Gibberly is already listening.
The flow in three steps
Step 1 — Audio capture. A laptop or tablet running Gibberly connects to a microphone (or receives audio via cable from the sound desk). The browser’s Web Speech API streams audio to Azure Speech AI.
Step 2 — Translation. Azure Cognitive Services translates the transcribed text in real time. Supported languages include German, English, Spanish, French, Portuguese, and dozens more.
Step 3 — Delivery. Translated text is pushed to every connected listener over Azure Web PubSub — a managed WebSocket service. Latency is typically under two seconds.
What listeners see
Congregation members scan a QR code displayed on screen (or printed in the bulletin). This opens a simple webpage in their browser. No app download. No account. Just their phone and a QR code.
Open source
Gibberly is open source under AGPL-3.0. The full source is available at github.com/ccoles146/gibberly. Self-hosters run it on their own Azure account — Microsoft’s nonprofit grant often covers the cost entirely.