Two safeguards. Every detection carries a visible confidence score in the operator panel — low-confidence hits never auto-go-live, they sit in a queue for the operator to approve or skip. And the operator always has a one-click "clear overlay" to pull anything off the broadcast instantly. The default posture is "surface, don't override": Kairos suggests, the operator decides.

Kairos ships with a local Whisper model that runs entirely on the booth machine — no internet required for transcription or detection. If you use cloud STT (Deepgram, Gladia, or our managed option) and the connection drops, Kairos falls back to the local model automatically so the broadcast keeps moving.

Typical end-to-end latency from the spoken word to the verse on the broadcast is 0.6 seconds. The pastor will not feel it; the congregation perceives it as the verse appearing as it's being preached, not after. The hardest case — a paraphrased quote — takes a beat longer because the semantic match runs after the phrase completes.

Live transcription audio is streamed to the STT provider you choose and is not retained after transcription — that holds for BYOK Deepgram/Gladia, our managed option, and local Whisper (which never leaves your machine). Optional session recordings are saved to your own cloud account on Kairos Cloud, accessible only to operators you invite. You own the audio; we don't train on it.

Yes. Operators control Kairos from a single panel — Prev / Next / Send Live, with keyboard shortcuts and a queue of pending detections. Most teams hand it off to a new volunteer in under fifteen minutes. Setup happens once, on the booth machine, and survives reboots.

Kairos is a free desktop app that detects Bible verses in real time from live sermon audio. It listens to a sermon feed, transcribes speech as it happens, identifies scripture references (including paraphrased quotations), and renders broadcast-ready overlays via NDI for OBS, vMix, and other live production tools. Kairos is built with Tauri v2, a React frontend, and a Rust backend.

A computer running Windows or macOS, an audio feed from your sound system (system audio, NDI, or a USB mic), and a projector or display screen on the output side. Kairos works with your existing audio setup and NDI-compatible switcher. No specialised hardware required.

Kairos supports KJV, ESV, NIV, NKJV, NLT, and more. You can switch between translations on-the-fly from the operator panel, and each translation is stored locally in your app database for offline use.

Kairos is trained to recognise paraphrased scripture references, not just exact quotations. A semantic / vector model surfaces the intended passage even when the wording differs from any source translation, and tags the match with a confidence score so the operator can decide whether to go live.

Yes, but their role becomes simpler. Instead of typing verse references into a search bar mid-sermon, the operator triages incoming detections — approving high-confidence hits, overriding the few that need correction. The job moves from data entry to judgement.

Yes. Kairos outputs broadcast-ready overlays via NDI, which integrates natively with OBS Studio, vMix, Resolume, and other professional live production software. The overlay is a single NDI source you drop onto a layer above your existing scene.

Download the free desktop app for Windows or macOS, plug in your audio feed, and you're ready to go. Most teams are running in under five minutes. Full setup instructions live in-app and in the online documentation.

No. Kairos is designed for church teams of all technical skill levels. Setup is straightforward, and once running, it operates with minimal interaction during services.

Kairos is free for every church when you bring your own transcription key (Deepgram or Gladia) or run our bundled local Whisper model fully offline. No account required, no usage limits. If you'd rather not manage a transcription account, every new account gets 2 free hours of managed transcription to try it — no card required. After that you can stay on a prepaid wallet ($10 ≈ 20 hours) or move to a Pro annual plan ($200/year with 60 hours included every month). See the pricing page for details.