📝 Pre-recorded Zoom Transcription App

Automatically transcribe your Zoom meeting recordings with Deepgram’s Speech-to-Text API. This script monitors your local Zoom recordings folder and automatically generates high-quality transcripts once a meeting ends.

---

🚀 Features

• Automatic Folder Monitoring — Detects new Zoom recordings using watchdog.
• Smart Delay Handling — Waits until Zoom finishes writing recording.conf before starting transcription.
• High-Accuracy Transcription — Uses Deepgram’s nova-3 model with diarization and utterance formatting.
• Speaker Labeling and Timestamps — Includes precise timestamps and speaker IDs in the transcript.
• Simple Setup — Just point the script to your Zoom recordings folder.

---

🧰 Requirements

Install dependencies:

pip install deepgram watchdog

Set your Deepgram API key:

export DEEPGRAM_API_KEY="YOUR_DEEPGRAM_API_KEY"

---

🏗️ How It Works

1. Zoom creates a folder for each meeting you record locally.
2. Once the meeting ends, Zoom writes a file named recording.conf that contains metadata (audio/video filenames).
3. The script waits until this file is complete.
4. When ready, it reads the audio path and sends the file to Deepgram for transcription.
5. A transcript file is generated and saved locally.

---

📂 Folder Structure

deepgram-58/
│
├── main.py              # Main watcher and transcription script
├── transcript.txt         # Generated transcript (auto-created)
└── README.md              # Documentation

---

⚙️ Usage

Run the script and specify the folder where Zoom saves recordings:

python script.py /home/user/Documents/Zoom

By default, the script:

• Watches the specified folder for new Zoom meeting folders.
• Waits for recording.conf to appear and be populated.
• Asks before starting transcription.
• Saves the final transcript to transcript.txt.

Example output:

📂 New folder detected: /home/user/Documents/Zoom/2025-10-16 14.28.03 Meeting
🎧 Transcribing: audio1862420270.m4a
✅ Transcript saved to ./transcript.txt

---

🧩 Configuration

You can adjust:

• Timeout for waiting on recording.conf (default: 30s)
• Transcript output path
• Model settings (e.g., use nova-3-general, nova-3-medical)

---

🧪 Example Transcript

Speaker 0 [00:00:00.000 - 00:00:04.230]: Good morning, everyone. Let’s get started.
Speaker 1 [00:00:04.230 - 00:00:07.900]: Morning! Can you hear me clearly?
Speaker 0 [00:00:07.900 - 00:00:10.500]: Yes, loud and clear.