Real-Time Speech Transcription

This script uses OpenAI's Whisper model to transcribe speech from your microphone in real-time and type it into any window you choose. The application supports multiple languages and provides visual feedback during transcription.

Features

• Real-time speech transcription using OpenAI's Whisper model
• Multi-language support with interface localization
• Visual feedback during active transcription
• Optional detection logging
• Cross-platform support (Windows and Linux)
• Automatic language detection option

Requirements

• Python 3.7+
• OpenAI Whisper
• PyAudio
• NumPy
• Pynput
• Pillow (for icon creation)
• Tkinter (for GUI)

Platform-specific Requirements

• Windows: Pywin32 and Winshell (for desktop shortcut creation)
• Linux: portaudio19-dev (for PyAudio)

Installation

Windows Installation

1. Run the setup script:

python setup.py

Ubuntu/Linux Installation

1. Install system dependencies:

sudo apt-get update
sudo apt-get install -y portaudio19-dev python3-dev python3-tk

2. Run the setup script:

python3 setup.py

Usage

Interface Features

1. Language Selection:

   • Choose from multiple supported languages including:
     • English
     • Portuguese
     • Chinese
     • German
     • Spanish
     • French
     • Italian
     • Japanese
     • Korean
     • Russian
   • Auto-detect option available
   • Interface language updates automatically with selection

2. Detection Logging:

   • Toggle logging on/off using the checkbox
   • When enabled, shows all detection attempts and transcriptions
   • Logs are saved to whisper_detection_log.txt

3. Visual Feedback:

   • Interface changes color during active transcription
   • Returns to normal color when transcription is paused/stopped

4. Status Display:

   • Shows current status and model loading progress
   • Displays first-time download information
   • Indicates when ready for trigger phrase

Basic Operation

1. Start the application using the desktop shortcut or command line
2. Select your preferred transcription language
3. Enable/disable logging as needed
4. Say "Hey Jarvis" to activate transcription
5. Speak your text to be transcribed
6. Uses any mouse/keyboard input to stop current transcription

Stopping Transcription Program

You can stop the transcription program by:

• Clicking the Stop button
• Closing the application window

Notes

• First-time startup will download the Whisper model large (about 2.9GB)
• The trigger detection is language-agnostic for better recognition
• Transcription automatically pauses when keyboard/mouse is used
• Interface language changes affect all text elements including acknowledgments
• Log toggle affects only detection logging, not status messages

Troubleshooting

Ubuntu/Linux Issues

1. PyAudio Installation:

   sudo apt-get install -y portaudio19-dev python3-dev
   pip3 install pyaudio

2. Desktop Shortcut:

   chmod +x ~/Desktop/real-time-transcriber.desktop

Windows Issues

1. Icon Not Showing: Try running as administrator once
2. Desktop Shortcut: Verify all paths in the shortcut are correct