Minutes High

Minutes High is a cross-platform .NET MAUI desktop app that turns Microsoft Teams meeting transcripts into clean, well-structured meeting minutes — completely on-device using a lightweight local LLM. No cloud, no API keys, your transcripts never leave your machine.

Features

• On-device AI — Downloads a lightweight Qwen2.5-1.5B-Instruct model (~1.1 GB, Q4_K_M) on first run and runs inference locally via LLamaSharp (llama.cpp).
• Teams transcript aware — Understands both the WebVTT (.vtt) export and the plain-text "copy transcript" format, stripping timestamps and merging speaker turns.
• Structured minutes — Produces Markdown with Title, Date, Attendees, Agenda, Summary, Key Discussion Points, Decisions, and Action Items.
• Modern UI — Clean, card-based design with light/dark theme support.
• Private by design — Everything runs locally.

How it works

1. On startup the app downloads the local model (only once; cached afterwards).
2. Load a Teams transcript (.txt or .vtt).
3. Click Format with AI — the transcript is normalized and summarized into minutes.

Requirements

• Windows 10 (10.0.17763.0) or later
• .NET 10 SDK (for building from source)
• ~1.5 GB free disk space for the model
• First run requires an internet connection to download the model

Getting the app

Download the release

Grab the latest Windows build from the Releases page, unzip it, and run minutesHigh.exe.

Build from source

git clone https://github.com/freyWylfred/minutesHigh.git
cd minutesHigh
dotnet build minutesHigh/minutesHigh.csproj -f net10.0-windows10.0.19041.0

To publish a self-contained Windows executable:

dotnet publish minutesHigh/minutesHigh.csproj -c Release -f net10.0-windows10.0.19041.0 `
	-p:WindowsPackageType=None

Project structure

Running the tests

dotnet test minutesHigh.Tests/minutesHigh.Tests.csproj

Tech stack

• .NET 10 / .NET MAUI
• LLamaSharp (llama.cpp) with the CPU backend
• Qwen2.5-1.5B-Instruct (GGUF, Q4_K_M)
• xUnit

License

This project is licensed under the MIT License. See LICENSE for details.

Acknowledgements

• Qwen by Alibaba Cloud
• LLamaSharp