Architecture

AI Processing Pipeline

The AI Processing Pipeline is a set of cloud‑based services that automatically transcribe sermons so people can search, find, and access sermon content.

What It Does

After a sermon is uploaded through the Admin Tool, the AI pipeline automatically:

1. Transcribes Audio — Converts sermon audio into searchable text
2. Generates Series Summaries (when series is complete) — Summarizes completed series

How It Works

   Staff uploads sermon
   via Admin Tool
           |
           v
   +------------------+
   | AI Processing    |
   | Pipeline         |
   +--------+---------+
            |
   Transcript stored
   (searchable, archived)
            |
            v
   Mobile App & Website
   (people search & find)

Processing Stages

The pipeline orchestrates these processing stages automatically:

1. Transcription Stage — Converts audio to searchable text.
2. Series Processing Stage (conditional) — When a series is complete, synthesizes all message content into a series summary.

Technical Foundation

The pipeline runs on a cloud-based, serverless architecture:

• Automatically triggered when sermons are uploaded or updated
• Scalable — handles demand without manual server management
• Integrated with the Global API database to read and update sermon records
• Distributed — processes audio, text, and metadata in parallel for speed

From a Practical View

From the staff and congregation’s perspective:

1. Staff uploads a sermon through the Admin Tool.
2. Pipeline runs automatically in the background—no manual steps needed.
3. Within minutes, the transcript is searchable and available.

The technology handles the complexity so your team can focus on ministry.

Why It Matters

This pipeline creates real value:

• Saves staff time — No manual transcription needed.
• Makes sermons discoverable — People can search for sermon content using text.

Why Cloud-Based?

The pipeline uses cloud services so it can:

• Scale automatically — Handle any number of sermons without manual intervention.
• Stay reliable — Retry automatically if something goes wrong.
• Improve over time — Get better as AI models improve, without disrupting your workflow.

This site is open source. Improve this page.