🎤 STT (Speech-To-Text) - Speech Recognition Projects

Comparison and exploration of different approaches for automatic speech recognition with smart punctuation.

Overview

This repository contains two implementations of speech-to-text systems:

🏗️ Approach	🔧 Technology	🎯 Advantage	⚡ Performance
C# WPF Desktop	Vosk + Microsoft.ML	Native Windows app	⭐⭐⭐
Python Web	Whisper + Flask	Automatic punctuation	⭐⭐⭐⭐⭐

Project Structure

STT/
├── README.md                    # 👈 This file
├── VoiceNoteAppByLud/          # 🖥️ C# WPF Application
│   ├── MainWindow.xaml
│   ├── MainWindow.xaml.cs
│   ├── App.xaml
│   ├── VoiceNotesApp.csproj
│   └── Models/
│       ├── vosk-model-fr/      # French Vosk model
│       └── encoder_model.onnx   # T5 for summarization
└── stt-whisper-py/             # 🌐 Python Web App
    ├── app.py                  # Flask + SocketIO
    ├── requirements.txt
    ├── static/                 # Web interface
    ├── templates/
    └── utils/                  # AI for summarization & classification

Approach Comparison

C# WPF Version (VoiceNoteAppByLud/)

Advantages:

• High-performance native Windows application
• Modern WPF interface with XAML
• No network dependency
• Complete system integration

Disadvantages:

• No automatic punctuation (Vosk limitation)
• .NET dependency complexity
• Windows-only
• Complex T5 tokenizer configuration

Recommended for: Enterprise Windows applications, offline environments

---

Python Web Version (stt-whisper-py/)

Advantages:

• Perfect automatic punctuation (Whisper)
• Responsive web interface (mobile/desktop)
• Real-time via WebSocket
• Automatic summarization + classification
• Cross-platform (Windows/Mac/Linux)

Disadvantages:

• Requires Python server
• More resource-intensive
• Local network dependency

Recommended for: Modern applications, distributed teams, rapid prototyping

Quick Start

Option 1: Python Web (Recommended)

cd stt-whisper-py/
pip install -r requirements.txt
python app.py
# → http://localhost:5000

Option 2: C# Desktop

cd VoiceNoteAppByLud/
dotnet build
dotnet run

Use Cases

Meeting Notes

• ✅ Python: Transcription + automatic summarization + classification
• ⚠️ C#: Raw transcription, manual summarization needed

Interviews

• ✅ Python: Perfect punctuation, JSON export
• ⚠️ C#: Correct transcription but without punctuation

Courses & Training

• ✅ Python: Automatic "course" classification, smart summarization
• ✅ C#: Stable local application, no network latency

Enterprise Environment

• ✅ C#: Local security, no data on server
• ✅ Python: Modern interface, collaboration possible

Technologies Used

Component	C# WPF	Python Web
STT Engine	Vosk	OpenAI Whisper
Interface	WPF/XAML	HTML5/CSS3/JS
Backend	.NET 8	Flask + SocketIO
Summarization	T5 ONNX	Transformers (mT5)
Classification	❌	LightGBM
Audio	NAudio	Web Audio API
Real-time	❌	WebSocket

Performance Metrics

Criteria	C# WPF	Python Web	Winner
Punctuation	❌	✅ Perfect	Python
Speed	✅ Fast	⚡ Real-time	Tie
Accuracy	⭐⭐⭐	⭐⭐⭐⭐⭐	🐍 Python
Setup Ease	⚠️ Complex	✅ Simple	🐍 Python
Stability	✅ Robust	✅ Robust	🤝 Tie
Security	✅ Local	⚠️ Server	🖥️ C#

Recommendations

Choose Python Web if:

• You want automatic punctuation (key feature)
• Need modern and responsive interface
• Require automatic summarization and classification
• Distributed team or collaboration needed
• Rapid prototyping

Choose C# WPF if:

• Maximum security required (local data only)
• 100% Windows environment
• Traditional enterprise application
• Critical native performance
• No server allowed

Future Development

Upcoming Python Features:

• Extended multilingual support
• REST API for integration
• Offline mode with local Whisper
• Advanced analytics dashboard

Possible C# Improvements:

• Azure Speech Services integration (punctuation)
• Whisper.NET plugin
• More modern interface (WinUI 3)

🤝 Contributing

Pull Requests are welcome for both projects!

• Issues: Report bugs and suggestions
• Features: Propose new features
• Docs: Improve documentation

📄 License

MIT License - Feel free to use for personal and commercial projects.

---

🎤 Happy transcribing! - May the AI force be with your words.