Live scripture projection

This is a submission for the DEV Weekend Challenge: Community

The Community
Every Sunday, I sit behind a laptop at the back of my church. I am part of the technical team, and one of my main jobs is to listen closely to the pastor, quickly search for the Bible verse they mention, and project it onto the screen for the congregation.

It sounds simple, but it is actually quite stressful. It is a lot of manual labor. You have to be incredibly fast, and when you are rushing to type, you are prone to errors. Sometimes you mishear the verse, or you make a typo, and the whole church is waiting on you. The friction of that process made me wonder what I was doing wrong, and it ultimately inspired me to build a solution for church technical teams everywhere.

What I Built

I built Rhema. Rhema is an application designed to take the stress and manual labor out of live scripture projection.

Instead of the media team scrambling to type, the application actively listens to the speaker and automatically detects when a Bible verse is mentioned. At this point, I have built a working core that proves the concept. The current features include:

— Real-time speech detection that listens for scripture references.
— Automatic verse extraction and display.
— Customizations in the settings to tweak how the detected verse is handled.

It has a lot of potential, but there is still plenty of work to do. My next steps are to implement a lot of UI fixes and sharpen the speech-to-text accuracy. Ideally, I would have used a paid AI API for the extraction pipeline right now, but I have no money for credits. Because of that constraint, finding creative, cost-effective ways to improve the speech parsing is next on my list of improvements.

Demo

I have created a test demo video to show exactly how the verse detection works in real-time.

Code

Code

live demo

How I Built It

Building real-time audio applications is always a challenge. I had to focus on getting a working core that people could actually see in action, rather than aiming for feature-completion on day one.

Because I couldn't lean on expensive, paid AI endpoints, I had to be creative with how the application parses the text stream. I relied heavily on native browser utilities—specifically the Web Speech API—for the transcription processing. To make it work, the logic involves taking that continuous raw transcript and passing it through a matching system to instantly extract book names, chapters, and verse numbers on the fly.

Switching to a mindset where I had to build a custom matching foundation instead of just querying an AI endpoint felt slow at first, but it forced me to keep the software simple, clean, and functional. The goal of software is to be used, and even in its current state, Rhema successfully removes a massive layer of friction for the person sitting behind the projection laptop.