🎙️ Whisper Realtime Echo 🚀

A modern GUI for Faster Whisper speech recognition model using Flet framework.

⚠️ This application has been tested and confirmed working on Windows. While it may work on other platforms (Linux, macOS), they have not been officially tested.

✨ Features

• 🔊 Transcribe audio files with a beautiful UI
• 🎤 Record audio directly and transcribe it
• 🎧 Live transcription with real-time speech recognition
• 🌍 Choose between English-only and multilingual models
• 🗣️ Select language for multilingual models (auto-detect available)
• 🌐 Translate any language to English text
• 📏 Select model size with VRAM/speed indicators
• 💻 GPU (CUDA) or CPU processing
• 📊 Real-time transcription status
• 🧩 Modular, maintainable codebase
• 📋 Copy results to clipboard with history access
• 🌓 Toggle between light and dark themes

📋 Model Information

Size	Parameters	English-only	Multilingual	VRAM	Speed
tiny	39 M	✅ tiny.en	✅ tiny	~1 GB	~10x
base	74 M	✅ base.en	✅ base	~1 GB	~7x
small	244 M	✅ small.en	✅ small	~2 GB	~4x
medium	769 M	✅ medium.en	✅ medium	~5 GB	~2x
large	1550 M	❌ N/A	✅ large	~10 GB	1x
turbo	809 M	❌ N/A	✅ turbo	~6 GB	~8x

💡 English-only models typically perform better for English transcription.

🚀 Included Pre-quantized Models

For optimal real-time transcription speed, the application includes pre-quantized int8 versions of the tiny models:

• int8_tiny_en: Pre-quantized tiny.en model for English transcription
• int8_tiny: Pre-quantized tiny model for multilingual transcription

These models were created using CTranslate2's conversion tool with the following command:

ct2-transformers-converter --model openai/whisper-tiny.en --output_dir int8_tiny_en --copy_files tokenizer.json preprocessor_config.json --quantization int8

🚀 Quick Start

Installation

# Clone repository
git clone https://github.com/alouiadel/whisper-realtime-echo.git
cd whisper-realtime-echo

# Create and activate virtual environment
# For Windows:
python -m venv venv
venv\Scripts\activate

# For macOS/Linux:
python -m venv venv
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

🔥 For GPU acceleration: If you want to use CUDA for GPU acceleration, install the appropriate PyTorch version from PyTorch official installation guide

Usage

# Run the application
python main.py

1. 📂 Click "Select Audio File" or 🎤 "Start Recording"
2. 🌐 Choose model type and size
3. 📊 Select processing device (CPU/GPU)
4. ▶️ Click "Transcribe" for file processing or 🎙️ "Start Live" for real-time transcription
5. 📝 View results and use the toolbar to:
   • 📋 Copy to clipboard
   • 📜 Access transcript history
   • 🌓 Toggle theme (top-right corner)

🎙️ Live Transcription:

• Provides real-time speech-to-text from your microphone
• Automatically detects and processes speech in chunks
• Optimized for maximum responsiveness:
  • Uses int8 quantization (reduces memory usage by ~40%)
  • Reduced beam size (1 instead of 5)
  • Greedy decoding with temperature=0
  • Independent chunk processing
  • Disabled timestamps generation
• Best practices:
  • Works best in quiet environments
  • Smaller models offer faster response times
  • Use a clear, consistent speaking voice

🎧 Supported Audio Formats

Whisper leverages ffmpeg to process audio, supporting a wide range of formats:

• Audio files: wav, mp3, m4a, ogg, flac, opus, amr
• Video files: mp4 (and other video formats with audio tracks)

📜 Clipboard History

• View, copy, and reuse previous transcriptions
• Saved between sessions with timestamps and model info
• Most recent 50 entries are retained

📝 Notes & Limitations

• Language Selection: Helps Whisper optimize recognition but doesn't translate content. English audio will still be transcribed as English even when another language is selected.

• Transcribe vs. Translate Modes:

  • Transcribe: Converts speech to text in the original language (default)
  • Translate: Converts speech from any language to English text

• Translation Best Practices:

  • Automatically sets language to "Auto-detect" for optimal results
  • Works best with larger models (large or turbo) and clear audio
  • Some languages may require transcribing first, then using a separate translation service

• System Requirements: If crashes occur, try using a smaller model or ensure your system meets the memory requirements in the model table.

🙏 Credits

This project is powered by Faster Whisper, a highly optimized implementation of OpenAI's Whisper. The Whisper model was trained on a large dataset of diverse audio and is capable of multilingual speech recognition, translation, and language identification.

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.