MeetingScriber

A real-time speech-to-text system that transcribes audio using OpenAI's Whisper model locally. The system consists of three microservices that work together to capture, process, and display transcriptions in real-time.

Architecture

The system consists of three main services:

1. Broadcaster: Captures microphone audio and streams it over WebSocket
2. Server: Processes audio using Whisper and broadcasts transcriptions
3. Client: Displays real-time transcriptions from connected broadcasters

Prerequisites

• Python 3.9 or higher
• A microphone for audio capture
• Operating System: macOS, Windows, or Linux

Installation and Setup

1. Clone the Repository

git clone https://github.com/encryptedtouhid/MeetingScriber.git
cd MeetingScriber

2. Set up Python Environment

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

# Install components based on your needs:

# For all components
pip install -r requirements.txt

# For server only
pip install -r server/requirements.txt

# For client only
pip install -r client/requirements.txt

# For broadcaster only
pip install -r broadcaster/requirements.txt

3. Install the Broadcaster Service

macOS

sudo ./install_scripts/macos/install_broadcaster.sh

Linux

sudo ./install_scripts/linux/install_broadcaster.sh

Windows

.\install_scripts\windows\install_broadcaster.bat

Running the Services

1. Start the Server

python server/main.py

2. Start the Client

python client/main.py

Running Broadcaster Directly (Without Service)

To run the broadcaster as a regular Python program instead of a system service:

python broadcaster/main.py

Note: When running the broadcaster directly, you'll need to keep the terminal window open. The service installation method is recommended for long-term use as it runs in the background and starts automatically with your system.

Project Structure

├── broadcaster/           # Audio capture service
│   ├── main.py           # Main broadcaster logic
│   └── service/          # Platform-specific service files
├── server/               # Whisper transcription service
│   └── main.py          # WebSocket server & transcription
├── client/              # Transcription display service
│   └── main.py         # Client UI
├── requirements.txt     # Python dependencies
└── install_scripts/     # Platform-specific installers

Configuration

Default WebSocket connections:

• Broadcaster → Server: ws://localhost:8000/audio
• Server → Client: ws://localhost:8000/transcribe

Development

1. Install development dependencies:

pip install -r requirements-dev.txt

2. Run tests:

pytest

Uninstallation

To uninstall the broadcaster service:

macOS

sudo ./install_scripts/macos/uninstall_broadcaster.sh

Linux

sudo ./install_scripts/linux/uninstall_broadcaster.sh

Windows

.\install_scripts\windows\uninstall_broadcaster.ps1

Troubleshooting

SSL Certificate Issues (macOS)

If you encounter SSL certificate errors when downloading the Whisper model, try:

pip install --upgrade certifi
export SSL_CERT_FILE="$(python -m certifi)"

WebSocket Connection Issues

• Ensure the server is running before starting the client
• Check if ports are not blocked by firewall
• Default ports can be modified in each service's configuration

License

MIT License