███╗ ███╗██╗████████╗██╗ ██╗██████╗ ██╗██╗
████╗ ████║██║╚══██╔══╝██║ ██║██╔══██╗██║██║
██╔████╔██║██║ ██║ ███████║██████╔╝██║██║
██║╚██╔╝██║██║ ██║ ██╔══██║██╔══██╗██║██║
██║ ╚═╝ ██║██║ ██║ ██║ ██║██║ ██║██║███████╗
╚═╝ ╚═╝╚═╝ ╚═╝ ╚═╝ ╚═╝╚═╝ ╚═╝╚═╝╚══════╝
██╗ ██╗██╗ ██╗██╗███████╗██████╗ ███████╗██████╗
██║ ██║██║ ██║██║██╔════╝██╔══██╗██╔════╝██╔══██╗
██║ █╗ ██║███████║██║███████╗██████╔╝█████╗ ██████╔╝
██║███╗██║██╔══██║██║╚════██║██╔═══╝ ██╔══╝ ██╔══██╗
╚███╔███╔╝██║ ██║██║███████║██║ ███████╗██║ ██║
╚══╝╚══╝ ╚═╝ ╚═╝╚═╝╚══════╝╚═╝ ╚══════╝╚═╝ ╚═╝
Mithril Whisper is a privacy-focused, air-gapped voice-to-text platform designed for organizations and individuals who prioritize data security. Built for environments where sensitive information cannot leave local systems, it provides enterprise-grade speech recognition with military-level security.
Mithril Whisper transforms speech into text using advanced AI models that run entirely on your local machine. No internet connection required. No data ever leaves your device. Perfect for government agencies, contractors, and privacy-conscious users who need powerful voice recognition without compromising security.
- Real-time Voice Transcription: Convert speech to text as you speak
- Global Hotkey Integration: Record from anywhere in your system with customizable key combinations
- Smart Text Cleanup: Automatic removal of filler words and intelligent formatting
- Auto-Injection: Seamlessly insert transcribed text into any application
- 100% Local Processing: All audio processing happens on your device
- Flexible Hotkeys: Use any key combination (F6, Ctrl+A, Alt+Space, etc.)
- Air-Gapped Operation: Works completely offline
- Zero Data Transmission: Audio never leaves your computer
- Bundled AI Models: Whisper.cpp models included - no downloads required
- Enterprise-Grade Security: Designed for classified and sensitive environments
- Transparent Processing: Open source - audit every line of code
- Immediate File Deletion: Audio files automatically deleted after transcription
In today's digital landscape, voice recognition has become essential for productivity, but most solutions require sending your voice data to cloud servers. Mithril Whisper solves this fundamental privacy problem by bringing enterprise-grade speech recognition directly to your desktop.
Whether you're working with:
- Classified Information - Government and military applications
- Proprietary Code - Software development in secure environments
- Confidential Documents - Legal, medical, and financial sectors
- Personal Privacy - Anyone who values data sovereignty
Mithril Whisper ensures your voice data remains under your complete control.
Navigate to the appropriate folder for your operating system:
Streamlined Windows implementation with:
- 100% local processing - no sign-in required
- Bundled Whisper.cpp models and binaries
- Windows-optimized audio processing
- Flexible hotkey system
Complete macOS implementation featuring:
- Native macOS integration
- Optimized for Apple Silicon and Intel Macs
- Seamless system-wide voice recognition
Mithril Whisper offers two distinct usage modes to meet different privacy and deployment needs:
Run the application directly from source code without any installation:
Privacy Features:
- ✅ No Authentication Required: Works immediately without login
- ✅ Zero Telemetry: No data collection or transmission whatsoever
- ✅ Complete Offline Operation: Never connects to internet
- ✅ Voice Data Never Stored: Audio files deleted immediately after transcription
- ✅ Local AI Processing: All transcription happens on your device using bundled Whisper.cpp
Windows Setup: Clone repository → cd mithril-whisper-windows → npm install → npm run dev
macOS Setup: Clone repository → cd mithril-whisper-mac → npm install → npm run dev
Pre-built applications for easy installation and updates:
Authentication:
- Requires user login for application access
- Login credentials managed through secure Supabase authentication
Telemetry (AI Assistant Only - macOS version):
- Windows Version: Zero telemetry - AI assistant removed for streamlined operation
- macOS Voice Dictation: Zero telemetry - completely private
- macOS AI Assistant Feature: Minimal telemetry when actively used
- Character count of text sent to/from AI assistant
- Session duration and timestamps
- Model used (e.g., whisper-tiny)
- Platform information (macOS)
- Content is NEVER tracked - only statistical metadata
What is NOT Tracked:
- Voice recordings or audio data
- Transcription content or text
- Conversations with AI assistant
- Personal or sensitive information
- Usage outside of AI assistant feature
Our telemetry implementation is fully transparent. Here's the actual code showing what data is collected (macOS version only):
// From mithril-whisper-mac/src/main/main.js - AI Assistant usage tracking only
const payload = {
transcript_chars_original: userPrompt.length, // Character count only
transcript_chars_cleaned: assistantReply.length, // Character count only
metadata: {
user_words: countWords(userPrompt), // Word count only
assistant_words: countWords(assistantReply), // Word count only
}
// NOTE: Actual content is NEVER included in telemetry
// NOTE: Windows version has NO telemetry - this code removed
};- Voice Input → Local Whisper.cpp processing → Transcribed Text (100% local)
- AI Assistant (macOS only, optional) → Character count metadata → Encrypted transmission to OpenAI
- File Cleanup → Audio files immediately deleted after processing
Bottom Line: Your voice never leaves your device. Windows version has zero telemetry. macOS version only tracks minimal usage statistics if you use the AI assistant feature (character counts, not content).
Mithril Whisper is part of the larger Mithril ecosystem - our comprehensive suite of open-source, air-gapped, zero-trust AI solutions. Designed for organizations with stringent security requirements, Mithril enables productive use of cutting-edge AI technologies without compromising data security.
- Voice Recognition (Mithril Whisper) - This repository
- Code Assistants - AI-powered development tools that work offline
- Document Analysis - Secure text processing and analysis
- Custom AI Deployments - Tailored solutions for specific organizational needs
With minimal hardware investment ($1,000-$2,000), organizations can deploy high-quality local AI models that never connect to the internet, providing all the benefits of modern AI in completely isolated environments.
Deploy Forward specializes in building secure AI systems for government agencies, defense contractors, and enterprises requiring the highest levels of data protection. We understand that in sensitive environments, the power of AI must be balanced with absolute security.
To democratize access to advanced AI capabilities while maintaining complete data sovereignty and security. We believe organizations shouldn't have to choose between cutting-edge technology and data protection.
- Security by Design: Every solution built with zero-trust principles
- Open Source Transparency: Full code visibility for security auditing
- Local Processing: AI that works without internet connectivity
- Custom Solutions: Tailored implementations for unique requirements
Ready to implement secure AI in your organization? We provide:
- Consultation: Assessment of your AI and security needs
- Custom Development: Bespoke solutions for your environment
- Training & Support: Comprehensive onboarding for your team
- Ongoing Maintenance: Long-term partnership for success
Josh Berns - Security-Focused AI Developer
Passionate about building AI solutions that respect privacy and maintain security. Specializing in air-gapped AI systems for sensitive environments where data protection is paramount.
- Email: boshjerns@gmail.com
- GitHub: @boshjerns
- Twitter: @boshjerns
- Company: Deploy Forward
Windows (Streamlined - No Assistant):
git clone https://github.com/boshjerns/MithrilWhisperApp.git
cd MithrilWhisperApp/mithril-whisper-windows
npm install
npm run dev # No authentication required, zero telemetry, no AI assistantmacOS (Full Features):
git clone https://github.com/boshjerns/MithrilWhisperApp.git
cd MithrilWhisperApp/mithril-whisper-mac
npm install
npm run dev # No authentication required, zero telemetry for transcriptionOptional AI Assistant Setup (macOS only):
# Add OpenAI API key for enhanced text processing (macOS only)
cd mithril-whisper-mac
cp env.example .env
# Edit .env and add: OPENAI_API_KEY=your_key_here- Choose Your Platform: Navigate to the appropriate platform folder
- Follow Detailed Setup: Each platform has comprehensive documentation
- Start Transcribing: Begin using voice recognition immediately
- Enhance (Optional - macOS only): Add OpenAI integration for advanced text processing
Platform Differences:
- Windows: Streamlined for pure voice transcription - no AI assistant, maximum privacy
- macOS: Full feature set including optional AI assistant with OpenAI integration
Mithril Whisper is available under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License for personal use, with commercial licensing available. We believe security through obscurity is a flawed approach - true security comes from transparency, peer review, and community validation.
We welcome contributions from the security and AI communities. Whether you're fixing bugs, adding features, or improving documentation, your input helps make voice recognition more secure for everyone.
As AI becomes increasingly central to productivity, the need for secure, local AI solutions grows. Mithril Whisper represents our vision of a future where powerful AI capabilities and complete data privacy coexist.
Join us in building a more secure, privacy-respecting AI ecosystem.
Built with ❤️ for organizations that value both innovation and security
Deploy Forward - Securing AI for Tomorrow