Meeting: local real-time transcription and generation of meeting minutes of the open source client

General Introduction

Meeting Minutes (a.k.a. Meetily) is a free and open source AI meeting assistant tool developed by Zackriya Solutions that focuses on capturing meeting audio in real-time, generating transcribed text, and automatically extracting meeting summaries. The tool runs entirely on local devices, supports macOS and Windows (Linux support is in development), and requires no external servers to ensure the privacy and security of user data. It is compatible with major meeting platforms such as Google Meet, Zoom, and Microsoft Teams, and has a built-in Whisper.cpp transcription engine and multiple Large Language Model (LLM) support for team collaboration and professional users who need to protect sensitive information. As of March 2025, the latest version is v0.0.3 pre-release, has received 1000+ stars on GitHub, and is licensed under the MIT license.

Function List

• Real-time audio capture: Supports capturing conference content from microphone and system audio for a wide range of conferencing platforms.
• real-time transcription: Converts audio to text from a natively running Whisper.cpp model, with support for multiple model sizes (tiny to large).
• Automatic summary generation: Utilizing LLMs (e.g. Anthropic, Groq, or local Ollama (Modeling) Analyze transcribed content to extract key points, action items, and decisions.
• local processing: All data processing is done on the user's device without the need for an Internet connection, guaranteeing privacy.
• semantic search: Intelligent retrieval of historical conference content is supported by ChromaDB vector storage and knowledge graph.
• Cross-platform support: Packaged apps are available for macOS and Windows, with Linux support coming soon.
• Export function (under development): Support for exporting transcriptions and summaries to Markdown, PDF, and HTML formats.
• modernized interface: Provides a responsive UI with real-time updates of transcription and summary content.

Using Help

Installation process

Meeting Minutes is a self-hosted tool that requires the user to complete the front-end and back-end installation on a local device. Below are the detailed steps based on the v0.0.3 pre-release version:

pre-conditions

• operating system: macOS or Windows (Linux support in development).
• software dependency::
  • forward part of sth.: No additional dependencies are needed, just use the pre-compiled installer.
  • back end::
    • Python 3.10 through 3.12.
    • FFmpeg (audio processing).
    • CMake and C++ compiler (build Whisper.cpp).
    • Windows users need to install Visual Studio Build Tools (with C++ development workloads).
• hardware requirement: 8GB of RAM or more recommended, multi-core processor, and support for GPU acceleration for improved performance.

Installation steps

1. Front-end installation

The front-end provides pre-compiled installation packages for Windows and macOS, which can be downloaded and installed directly.

Windows Installation

• Download the installation package: Access v0.0.3 Releases page, select any of the following documents:
  • Option 1 (recommended)::meeting-minutes-frontend_0.1.0_x64-setup.exeThe
  • Option 2::meeting-minutes-frontend_0.1.0_x64_en-US.msiThe
• Run the installation program::
  1. Double-click the downloaded .exe maybe .msi Documentation.
  2. Windows will bring up a "Windows Protect Your Computer" prompt that says "Microsoft Defender SmartScreen is blocking unrecognized applications":
     • Click on "More Information".
     • Select "Still running".
  3. If it still doesn't run, right-click on the installation file:
     • Select Properties.
     • Check the "Unlock" checkbox at the bottom and click "OK".
  4. Follow the installation wizard to complete the installation and the application will appear on the desktop.
• Delegation of authority: Allow microphone and audio capture permissions on first run.

macOS Installation

• Download the installation package: from v0.0.3 Releases page downloading dmg_darwin_arch64.zipThe
• Installation steps::
  1. decompression (in digital technology) dmg_darwin_arch64.zipget meeting-minutes-frontend_0.1.0_aarch64.dmgThe
  2. double-click .dmg Documentation.
  3. commander-in-chief (military) meeting-minutes-frontend.app Drag and drop into the Applications folder.
  4. Run the following command in a terminal to remove the quarantine attribute:

     xattr -c /Applications/meeting-minutes-frontend.app
     

• Delegation of authority: Allow microphone and audio capture permissions on first run.

2. Back-end installation

The back-end needs to be deployed manually, here are the complete steps:

1. clone warehouse

git clone https://github.com/Zackriya-Solutions/meeting-minutes.git
cd meeting-minutes/backend

2. Installation of predecessor dependencies

• macOS::
  • Install FFmpeg via Homebrew:brew install ffmpegThe
  • Install CMake:brew install cmakeThe
• Windows (computer)::
  • Install FFmpeg with Chocolatey:choco install ffmpegThe
  • Install Visual Studio Build Tools (with C++ workloads).
  • Make sure Python 3.10+ is added to the PATH.

3. Configure environment variables (optional)
   If using Anthropic maybe Groq The LLM needs to be created .env file and add the API key:

• Get the key:
  • Anthropic: https://www.anthropic.com/.
  • Groq: https://console.groq.com/.
• Configuration:
  • macOS:

    echo -e "ANTHROPIC_API_KEY=your_api_key\nGROQ_API_KEY=your_groq_api_key" > .env
    

  • Windows (PowerShell):

    "ANTHROPIC_API_KEY=your_api_key`nGROQ_API_KEY=your_groq_api_key" | Out-File -FilePath .env -Encoding utf8
    

• Set the Groq environment variable:
  • macOS/Linux:

    export GROQ_API_KEY=your_groq_api_key
    

  • Windows (PowerShell):

    $env:GROQ_API_KEY="your_groq_api_key"
    

  • Windows (Command Prompt):

    set GROQ_API_KEY=your_groq_api_key
    

4. Building Whisper.cpp

• macOS:

chmod +x build_whisper.sh
. /build_whisper.sh

• Windows:

. \build_whisper.bat

(If the script fails, run the commands within the script manually)

5. Start the back-end service

• macOS:

. /clean_start_backend.sh

• Windows:

. \start_with_output.ps1

Usage

Once the installation is complete, users can use Meeting Minutes by following the steps below:

1. Real-time transcription of meetings

• launch an application: Double-click the front-end application on your desktop, or run the back-end service.
• Selecting the audio source: Select "Microphone" or "System Audio" in the interface.
• Start recording: Click on the "Start" button and the live transcription will be displayed on the interface.
• Stop & Save: Click "Stop" at the end of the meeting and the transcription is automatically saved locally.

2. Generation of meeting summaries

• Configuring LLM: Select the local model or external service in the setup (required) .env (Configuration). It is recommended to use models with 32B parameters or higher.
• Generate a summary: Click on "Generate Summary" when transcription is complete to view key points, action items and decisions.
• Save results: Currently saved as a local file, export function (Markdown/PDF) under development.

3. Semantic search

• View History: Click "History" to view the saved meetings.
• Search Content: Enter a keyword and the tool returns the relevant snippet.
• Jump to view: Click on the results for details.

operating skill

• Enhancing transcription quality: To use a clear microphone, select the large version of the Whisper model.
• optimize performance: If you don't have enough hardware, use the Groq API to speed things up.
• compatibility: Adjust the audio input device in the conference software.

caveat

• Known issues::
• Smaller LLMs (<32B parameters) have lower summary quality.
• CMake and C++ are required for the backend build, and initial configuration is complicated.
• be in favor of: Questions can be answered through the Discord or GitHub Feedback.