Mityu: Privacy-First Local AI Meeting Assistant for macOS & Windows

Introduction

Mityu is a privacy-first, local-first AI meeting assistant built as a Tauri 2 desktop application. It combines a robust Rust core with a modern Next.js UI to provide comprehensive meeting and conversation intelligence. Mityu records meetings and on-site conversations, transcribes them directly on your device, and transforms them into structured, source-linked summaries and action items. Crucially, it operates with 100% local processing, requiring no cloud services or external servers, making it ideal for professionals and enterprises prioritizing privacy and control over sensitive data. Mityu is a product by bluedev.

Why Use Mityu and Key Benefits

Mityu stands out for its commitment to privacy, local processing, and flexibility, offering several compelling advantages:

Privacy-First: All capture, transcription, and summarization processes run entirely on your local device. No audio or transcripts are sent to external servers, ensuring your sensitive conversations remain private and secure.
Local-First Operation: The application functions completely offline, without any network or server requirements. An optional sync/collaboration server may be added in the future, but the core desktop app will always work with local data.
Cost-Smart: By leveraging local models, Mityu helps you avoid pay-per-minute bills associated with cloud-based AI services, offering a cost-effective solution for meeting intelligence.
Flexible Model Integration: You have the freedom to use local open-source models or plug in external API services. Mityu supports a "Bring Your Own Key" (BYOK) approach, preventing vendor lock-in.
Platform Agnostic Meeting Capture: Mityu captures system audio from any meeting application, including Google Meet, Zoom, or Teams, as well as in-person conversations. It is not a bot that joins your call. System audio capture is supported on macOS (ScreenCaptureKit / Core Audio tap) and Windows (WASAPI loopback). Linux support for system audio is experimental.
Human-in-the-Loop Summaries: AI-generated summaries and action items are presented as drafts, directly linked to their source transcript segments. This design ensures human approval is required, fostering trust, providing dispute evidence, and aligning with transparency requirements.
Advanced Transcription: Features local transcription using powerful models like Whisper (large-v3) or NVIDIA Parakeet, delivering real-time transcripts of your meetings.
Diverse AI Summarization Options: Choose from various AI models for summarization, including local Ollama, Claude, Groq, OpenRouter, or any OpenAI-compatible endpoint. API keys are securely stored in your OS keychain.
Audio Import and Enhancement: Import existing audio files to generate transcripts or re-transcribe recordings with different models or languages, all processed locally.
Professional Audio Mixing: Capture microphone and system audio simultaneously, with features like ducking and clipping prevention for clear recordings.
GPU Acceleration: Benefit from hardware acceleration on Apple Silicon (Metal) + CoreML on macOS, and NVIDIA (CUDA) and AMD/Intel (Vulkan) on Windows/Linux, enabled automatically at build time.
Multi-Platform Support: Mityu is available for macOS and Windows, with Linux builds supported from source.

Installation

Mityu provides straightforward installation processes for supported operating systems.

Windows

Download the latest x64-setup.exe from the Mityu Releases page.
Run the installer and follow the on-screen prompts.

macOS

Download the .dmg file from the Mityu Releases page.
Open the .dmg file and drag the Mityu application to your Applications folder.

Linux (Experimental)

Please note that Linux support is experimental. Recording, transcription, and summarization work, but system audio capture is not yet fully supported. Mityu will record the microphone only.

To build from source:

git clone https://github.com/aydogandagidir/mityu
cd mityu/frontend
pnpm install
./build-gpu.sh

For more detailed instructions, refer to the Building on Linux guide and the general build guide.

Examples

Mityu's user interface is designed for clarity and efficiency.

Source-linked AI summary: The application provides AI-generated summaries that are directly linked back to specific segments of the transcript, allowing for easy verification and context.
Microphone + system audio device selection: Users can easily select and configure their microphone and system audio input devices, ensuring optimal recording quality for various scenarios.