How It Works

Three stages — Listen, Understand, Display — run continuously while your mic is open.

1 · Listen

Your browser captures the microphone as raw 16 kHz audio and streams it to the Scrivent backend, which proxies it to a speech-to-text engine. Partial and final transcripts come back as you speak.

2 · Understand

Each phrase is matched against scripture using two strategies in parallel:

• Exact path — references like "John 3:16" and verbatim quotes are matched directly and instantly.
• Semantic path — paraphrases are embedded and compared against every verse by meaning, so "God gave his only son" still finds John 3:16.

Each candidate gets a confidence score; low-confidence matches are filtered out to avoid showing the wrong verse.

3 · Display

The best match is pushed over a live WebSocket to every connected display — projector, overlay, and NDI — typically in under two seconds end-to-end.