GCISD Watchdog

With GCISD making important decisions about school closures and rezoning, it's more crucial than ever to stay informed about what happens in Board of Trustees meetings. Because GCISD’s YouTube channel does not provide automatic transcriptions for these meetings, this project helps fill that gap by providing searchable, high-quality transcripts. GCISD Watchdog is a custom ChatGPT assistant that lets you ask questions about board meetings in plain English and get answers based on actual meeting transcripts.

Try the GPT:

• Search for "GCISD Watchdog" in the GPTs section of ChatGPT, or
• Use this direct link: GCISD Watchdog GPT

What's in this repository: This repo contains both the transcription tools and the raw transcripts themselves. You can browse individual meeting transcripts in the transcripts/ folder or use the tools below to generate your own dataset.

About Transcript Quality

Current transcripts are generated using the ggml-tiny.en-q5_1 model, which is the smallest quantized Whisper model available. This allows for fast processing but may result in some inaccuracies in the transcripts.

Our roadmap:

1. First, transcribe all available board meetings to ensure complete coverage - Done
2. Then, improve quality by re-transcribing with larger, more accurate models

Verification: The GCISD Watchdog GPT can link back to the original YouTube videos with exact timestamps for direct quotes, making it easy to verify statements in context. However, please note that the authors of this repository are not liable for incorrect transcripts or GPT hallucinations. Always verify important information against the official GCISD channel.

⚠️ Found a discrepancy?

If you spot a glaring discrepancy, error, or something that looks incorrect in any transcript, please help improve the dataset by opening an issue on GitHub:

👉 Open a transcript quality issue

Your reports help keep the transcripts accurate and trustworthy for the community. Thank you for contributing!

Running the Transcription Pipeline

This tool follows a simple process to create a searchable AI assistant from GCISD board meeting videos:

YouTube Videos → download → videos/*.webm → transcribe → transcripts/*.txt → assemble → gpt/dataset.md

Step 1: Build the Docker Image

First, build the container (this downloads the AI transcription model):

docker compose build

Step 2: Download Board Meeting Videos

Download videos from the GCISD YouTube channel:

docker compose run --rm download

This saves audio streams as .webm files in the videos/ folder.

Step 3: Transcribe Videos to Text

Convert the downloaded videos into text transcripts:

docker compose run --rm transcribe

This creates individual .txt transcript files in the transcripts/ folder.

Step 4: Assemble Dataset for ChatGPT

Combine all transcripts into a single dataset file:

docker compose run --rm assemble

This creates gpt/dataset.md with all meeting transcripts organized by date. Upload this file to ChatGPT to create your AI assistant that can answer questions about GCISD board meetings.