What audio does automatic verse detection need?

A clean feed carrying the preacher's microphone, ideally taken from the sound desk. The closer the audio is to the voice and the less room reverb it contains, the more reliably the software recognises references and paraphrases.

Can verse detection work from a laptop microphone?

It can, and it will be noticeably worse. A laptop microphone in a booth hears the room — reverb, congregation, air conditioning and the preacher somewhere in the distance. It is a reasonable way to try the software and a poor way to run it weekly.

Why does detection miss verses in our building?

Usually reverb or a distant microphone. Recognition depends on consonants, and a reverberant room smears exactly those. A desk feed from the preacher's own microphone bypasses the room entirely and is the single biggest improvement available.

Does the preacher need to speak differently?

No. The point of detection is that it works with normal preaching, including paraphrase and quotation from memory. Clear speech helps, as it helps the congregation, but nobody should be changing how they preach for the software.

Audio

Audio for verse detection: what the software needs to hear

By Mambo Mutinda · Founder & CEO, TajiCast

4 September 2026 · Updated 8 September 2026

Automatic verse detection is a speech problem before it is a Bible problem. Everything it can do depends on what reaches it, and the difference between a good feed and a bad one is larger than the difference between any two pieces of software.

Give it the desk feed

The single most important decision, and it is usually free.

A feed from the sound desk carries the preacher's microphone: close to the mouth, mixed, and largely free of the room. That is what recognition works best with.

A room microphone — a laptop microphone in the booth, or a camera microphone at the back — hears the building. Reverb, the congregation, air conditioning, someone's chair, and the preacher at a distance. The words are there for a human listener; they are considerably harder for recognition.

If you can route only one thing to the machine running detection, route the preacher's microphone.

This is the same feed that improves your stream, your recording and your transcripts. One cable, four benefits — which is why it is the first thing to fix in most church booths.

What makes recognition hard

Reverb. The main problem in most sanctuaries. Recognition depends heavily on consonants — the difference between "Philippians" and "Ephesians" is carried in short, quiet sounds — and reverb smears exactly those. A hard, reflective room degrades detection the same way it degrades intelligibility for people at the back.

Distance. Every doubling of distance from the mouth adds room and loses clarity. A lapel is further than a headset; a lectern microphone is further still; a room microphone is furthest.

Low level. Audio that arrives too quiet and is amplified later brings its noise floor up with it. Set the level properly at the input — see gain structure.

Clipping. Distorted audio is harder to recognise than quiet audio. If the loudest moments crackle, the level into the machine is too high.

Music underneath speech. A band playing quietly behind a closing exhortation makes recognition much harder. Nothing to fix, but worth knowing why detection goes quiet there.

ImageDiagram: the same sentence reaching a headset mic, a lectern mic and a room mic, with reverb and distance illustrated

Distance and reverb are what stand between the voice and the recogniser.

Setting it up

Take an output from the deskAn aux, matrix or subgroup carrying the preacher's microphone. The main output works if that is all you have.
Into the machineBy USB from the desk, or through a USB audio interface. Details in connecting a mixer.
Set a healthy levelPeaks well short of clipping, with the preacher speaking at real volume — not a soundcheck murmur.
Choose the input in the softwarePoint detection at that device rather than at the default microphone, which is usually the laptop's own.

That last step catches people: the software works, badly, on the built-in microphone, and nobody realises the desk feed was never selected.

Testing it properly

Use your own preacher. A demonstration clip of clear studio speech tells you nothing about your building. Play back a recording of a recent sermon — your room, your microphone, your preacher's pace and accent — and watch what detection does with it.

Test the hard parts deliberately:

A verse quoted from memory, without a reference.
A chain of cross-references a sentence apart.
A jump to a different book.
A passage read while the band plays quietly.

Watch what it does when unsure, which matters as much as what it does when confident.

What not to change

Do not ask the preacher to speak differently. The point of detection is that it works with normal preaching, including paraphrase and quotation from memory. A preacher announcing every reference clearly for the software's benefit has been made to serve the tool.

Do not chase the last few percent. Detection that catches the great majority, with an operator who can correct the rest in a second, is a working system. Perfect recognition is not the goal; a screen that keeps up with the sermon is.

ImageScreenshot: detection running during a service, showing a recognised verse and the confidence the software has in it

What it does when it is unsure matters as much as when it is confident.

The takeaway

The desk feed is the whole game. A room microphone will work and will disappoint.
Reverb and distance are what degrade recognition, for the same reason they degrade a sermon at the back of the room.
Test with your own preacher in your own building, and check the awkward cases deliberately.

Give it a clean feed and let it keep up with the sermon. Free, and it runs offline.

Download TajiCast, free

Frequently asked questions

What audio does automatic verse detection need?: A clean feed carrying the preacher's microphone, ideally taken from the sound desk. The closer the audio is to the voice and the less room reverb it contains, the more reliably the software recognises references and paraphrases.
Can verse detection work from a laptop microphone?: It can, and it will be noticeably worse. A laptop microphone in a booth hears the room — reverb, congregation, air conditioning and the preacher somewhere in the distance. It is a reasonable way to try the software and a poor way to run it weekly.
Why does detection miss verses in our building?: Usually reverb or a distant microphone. Recognition depends on consonants, and a reverberant room smears exactly those. A desk feed from the preacher's own microphone bypasses the room entirely and is the single biggest improvement available.
Does the preacher need to speak differently?: No. The point of detection is that it works with normal preaching, including paraphrase and quotation from memory. Clear speech helps, as it helps the congregation, but nobody should be changing how they preach for the software.

Keep reading