Local Meeting Notes with Whisper Transcription + Ollama Summaries (Gemma3n, LLaMA, Mistral)

Introduction

Taking notes during meetings is one of those universal pain points. Most teams either assign someone to jot down minutes, or rely on cloud-based AI note takers like Otter, Fireflies, or Jamie. But these come with trade-offs:

Your data lives on someone else’s servers.
Subscription costs scale with team size.
Integrations can break or require permissions you’d rather not grant.

Meetily was built to solve this problem differently: a self-hosted, open-source desktop app that records, transcribes, and summarizes your meetings completely offline. With the release of v0.0.5, Meetily has taken a big step forward in stability, deployment, and local AI model support.

What Is Meetily?

Meetily is an AI-powered meeting companion that you run locally on your own machine. Unlike SaaS tools that join your calls as bots or plugins, Meetily:

Captures audio directly from your system output and microphone.
Works with any platform: Zoom, Teams, Google Meet, Discord, or even in-person recordings.
No integrations required — no bots, browser extensions, or special permissions.
Cross-platform: available on Windows, macOS, and Docker (with Linux support on the way).

In other words, Meetily is a standalone desktop app. Once installed, it just works — regardless of which conferencing software your team prefers.

Whisper for Local Transcription

At the heart of Meetily is Whisper, the speech recognition model from OpenAI. Instead of sending audio to a server for transcription, Meetily runs Whisper locally, turning your audio into accurate text right on your machine.

Benefits of local Whisper:

Privacy-first: no audio leaves your computer.
Faster feedback: transcription happens in real time or shortly after your call.
Cost savings: no API usage fees.

This makes Meetily especially useful for professionals working with sensitive information — consulting, law, healthcare, or research.

Ollama Summaries: Your Own LLM on Your Machine

Once Whisper has transcribed the conversation, Meetily passes the text to your local Ollama instance for summarization. Ollama acts as a lightweight, local LLM runtime that lets you run advanced models with a simple command.

Supported Models

Meetily works seamlessly with Ollama-supported models, including:

Gemma 3n — strong summarization capabilities with good performance.
LLaMA (Meta) — popular general-purpose models.
Mistral — efficient, lightweight models well-suited for fast local inference.

Because everything runs through Ollama:

No API keys required (unless you want to connect external providers).
Everything is local: your transcripts and summaries never leave your machine.
Customizable: choose the model that fits your hardware and use case.

What’s New in v0.0.5

This release focuses on making Meetily easier to deploy and faster to use:

Stable Docker support — run Meetily on x86_64 and ARM64 environments reliably.
Native installers — Windows and macOS builds (with Homebrew tap support for Mac users).
Streamlined setup — simplified dependency management and installation.
Backend optimizations — faster Whisper transcription and smoother Ollama summarization pipelines.

These updates make Meetily friendlier for both developers and end users.

Installation Guide (Aligned with Release Notes)

macOS — Install via Homebrew (Recommended)

brew tap zackriya-solutions/meetily
brew install --cask meetily

Launch Meetily from Applications or Spotlight.
On first run, grant microphone and screen recording access.
Start the backend server if required:

  meetily-server --language en --model medium

Upgrading from v0.0.4? Run:

  brew update
  brew upgrade --cask meetily
  brew upgrade meetily-backend

Data is automatically migrated to /opt/homebrew/var/meetily/meeting_minutes.db.

Windows — Installer

Step 1: Install the Frontend Application

Download the installer:
- Go to Latest Releases
- Look for and download the file ending with x64-setup.exe (e.g., meetily-frontend_0.0.5_x64-setup.exe)
Prepare the installer:
- Navigate to your Downloads folder
- Right-click the downloaded .exe file → Properties
- At the bottom, check the Unblock checkbox → Click OK
- This prevents Windows security warnings during installation
Run the installation:
- Double-click the installer to launch it
- If Windows shows a security warning:
  - Click More info → Run anyway, OR
  - Follow any permission dialog prompts
- Follow the installation wizard steps
- The app will be installed and available on your desktop/Start menu

Step 2: Install and Start the Backend

Download the backend:
- From the same releases page
- Download the backend zip file (e.g., meetily_backend.zip)
- Extract the zip to a folder like C:\meetily_backend\
Prepare backend files:
- Open PowerShell (search for it in Start menu)
- Navigate to your extracted backend folder:
```
 cd C:\meetily_backend
```

Start the backend services:

   .\start_with_output.ps1

This script will:
- Guide you through Whisper model selection (recommended: base or medium)
- Ask for language preference (default: English)
- Download the selected model automatically
- Start both Whisper server (port 8178) and Meeting app (port 5167)

What happens during startup:

Model Selection: Choose from tiny (fastest, basic accuracy) to large (slowest, best accuracy)
Language Setup: Select your preferred language for transcription
Auto-download: Selected models are downloaded automatically (~150MB to 1.5GB depending on model)
Service Launch: Both transcription and meeting services start automatically

✅ Success Verification:

Check services are running:
Test the application:
- Launch Meetily from desktop/Start menu
- Grant microphone permissions when prompted
- You should see the main interface ready to record meetings

Docker — Cross-Platform

For server environments or those who prefer containers:

Best for: Developers, testing, or when you want automatic dependency management
Performance Note: 20-30% slower than native installation

Quick Start (5-10 minutes)

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

# Windows (PowerShell)
.\build-docker.ps1 cpu
.\run-docker.ps1 start -Interactive

# macOS/Linux (Bash)
./build-docker.sh cpu
./run-docker.sh start --interactive

What you get:

Automatic dependency management
No need to install Python, CMake, or build tools
Consistent environment across all platforms
Interactive model selection
Both services running: http://localhost:8178 and http://localhost:5167

Configuring Ollama

Install Ollama: https://ollama.com/
Pull models:

   ollama pull gemma:latest
   ollama pull llama:latest
   ollama pull mistral:latest

In Meetily → Settings → LLM Provider, set:

Provider: Ollama (local)
Base URL: http://localhost:11434
Model: e.g. gemma:latest

Getting Started

Launch Meetily and select System Audio + Microphone.
Start recording → Whisper transcribes locally.
At the end of the meeting, Ollama summarizes into action items and key notes.
Review meeting history, stored locally on your machine.

Community & Support

Meetily is built in the open, and we welcome feedback, contributions, and bug reports.

Conclusion

With Whisper for transcription and Ollama for local LLM summaries, Meetily offers a fast, private, and fully self-hosted alternative to cloud meeting assistants. Version 0.0.5 brings Docker stability, native installers, and smoother pipelines — making it easier than ever to run AI meeting notes entirely on your own hardware.

If you’ve been waiting for a way to keep your meeting data private, flexible, and under your control, now’s the time to give Meetily a try.

👉 Download Meetily v0.0.5

Source link