Congressional Hearing Transcriber

A Flask-based web application that automatically transcribes congressional hearing audio files, identifies speakers, and generates analytical summaries. Built for researchers, journalists, and policy analysts working with congressional hearing recordings.

Features

• High-accuracy speech-to-text transcription using OpenAI Whisper
• Automatic speaker identification and separation using Pyannote (not yet attributing the actual identity of the speaker)
• Web interface for easy file uploads (not yet tested)
• Multiple output formats (JSON, TXT, SRT)
• Real-time processing status updates
• Secure file handling and storage (at your own risk)

Installation

1. Clone the repository: git clone https://github.com/YOUR-USERNAME/hearing-transcriber.git cd hearing-transcriber

2. Create and activate a virtual environment: python -m venv venv source venv/bin/activate # On Windows: venv\Scripts\activate

3. Install required packages: pip install -r requirements.txt

4. Create a .env file with required tokens: SECRET_KEY=your_flask_secret_key - This is a random string you generate for Flask security - You can generate one in Python using: import secrets print(secrets.token_hex(16) - Copy the output and use it as your SECRET_KEY

   HF_TOKEN=your_huggingface_token

   • Create account at Hugging Face (https://huggingface.co/)
   • Go to Settings → Access Tokens
   • Click "New token"
   • Give it a name (e.g., "hearing-transcriber")
   • Select "read" access
   • Accept the model terms:
   • Visit Speaker Diarization model
   • Click "Accept terms" button
   • Visit Segmentation model
   • Click "Accept terms" button

Prerequisites

1. System Requirements:

   • Python 3.10+
   • ffmpeg
   • 8GB+ RAM recommended

2. API Access:

   • Hugging Face account (https://huggingface.co)
   • Accept terms for pyannote/speaker-diarization
   • Accept terms for pyannote/segmentation

Usage

1. Start the Flask server: python run.py

2. Open your browser and navigate to: http://localhost:5000

3. Upload an audio file (MP3 or WAV format)

   • Maximum file size: 16MB
   • Supported formats: MP3, WAV

4. Wait for processing to complete

   • Progress will be shown in real-time
   • Results will be available for download

Output Files

• transcript.json: Full transcript with speaker labels and timestamps
• transcript.txt: Human-readable transcript with summary
• transcript.srt: Subtitle format with speaker identification

Project Structure

/app /static - Static assets and uploaded images /storage - Temporary file storage /templates - HTML templates /transcriber - Core transcription logic /config - Application configuration

Security Notes

• All uploaded files are securely handled
• Temporary files are automatically cleaned
• File extensions are strictly validated
• Maximum file size is enforced

License

MIT License - See LICENSE file for details

Contributing

Contributions are welcome! Please read CONTRIBUTING.md for guidelines.