About This Tool

Understanding how AI transforms meeting recordings into actionable notes

🎯 Overview

The Meeting Notes Generator is a proof of concept application developed by the AI Expert Team at The Hague University of Applied Sciences (THUAS). It demonstrates how modern AI technologies can be combined to automate the tedious task of creating meeting notes from recordings.

⚙️ The Pipeline

📁

Upload

Video/Audio file

🔊

Extract Audio

FFmpeg processing

🎙️

Transcribe

ASR Model

🤖

Process

LLM Analysis

📝

Output

Structured Notes

🎙️ Speech Recognition Models

NVIDIA Parakeet

  • Model: nvidia/parakeet-tdt-0.6b-v2
  • Parameters: 600M
  • Languages: English
  • Speed: Fastest (~10x real-time)
  • Best for: Quick English transcriptions

Mistral Voxtral-Mini

  • Model: mistralai/Voxtral-Mini-3B-2507
  • Parameters: 3B
  • Languages: EN, NL, DE, FR, ES, PT, IT
  • Speed: Moderate (~3x real-time)
  • Best for: Multilingual meetings

🤖 LLM Models (via Ollama)

Ministral 3B

Fast reasoning with good accuracy. Best for quick results.

Ministral 8B

Balanced performance. Recommended for most use cases.

Ministral 14B

Best quality output. Use when accuracy is critical.

✨ Features

  • 📋
    Structured Output

    Summary, Topics, Decisions, Action Items, Risks, Open Questions

  • ⚙️
    Customizable Config

    Participants, glossary, meeting context for better accuracy

  • 📝
    Editable Prompts

    Customize how the LLM generates each section

  • 📥
    Multiple Formats

    Download as Markdown or DOCX

🛠️ Technology Stack

Backend

FastAPI, Python

ASR

NeMo, Transformers

LLM

Ollama (local)

Audio

FFmpeg, SciPy

Frontend

HTML, CSS, JS

Export

python-docx

🚀 Roadmap

Core Pipeline

Upload → Transcribe → Generate Notes

Multiple Models

Support for different ASR and LLM models

Speaker Diarization

Identify who said what

Real-time Processing

Live meeting transcription

Teams Integration

Direct integration with Microsoft Teams

👥 Credits

Developed by the AI Expert Team at The Hague University of Applied Sciences (THUAS)

This is a proof of concept application for research and demonstration purposes. All processing is done locally using open-source models.