Meetily Setup: Free Local AI Meeting Notes in 10 Minutes

Meetily is a free, open source AI meeting assistant that records, transcribes and summarizes your calls entirely on your own machine, and today it is the hottest repository on GitHub trending, adding roughly 2,500 stars in a single day on top of nearly 20,000 total. This is the first entry in our new daily Tutorials section: one trending GitHub tool, set up for real. Ten minutes takes you from download to a private, working replacement for cloud notetakers like Otter and Fireflies, with no subscription and no audio ever leaving your laptop.

• Meetily v0.4.0 ships one-click installers for Windows and macOS; Linux builds from source in four commands.
• Everything runs locally: Whisper or Nvidia Parakeet models transcribe on device, and Ollama generates the summaries.
• It captures microphone and system audio directly, so it works with Zoom, Meet or Teams without a bot joining your call.
• MIT licensed and free; a paid PRO edition exists, but the maintainers commit to keeping the community edition free and open source forever.

What is Meetily and why is it trending?

Meetily is a desktop meeting assistant from Zackriya Solutions, built as a Tauri app with a Rust backend and a Next.js interface. It records meeting audio, produces a live transcript with Whisper or Nvidia's Parakeet models, and generates structured summaries with a local language model through Ollama. The project claims its Parakeet pipeline transcribes around 4x faster than Whisper. Version 0.4.0 shipped on June 5, 2026, and the repository sits at nearly 20,000 stars under an MIT license. The trending spike has an obvious driver: meeting audio is the most sensitive data stream a company produces, and the README leans hard on that fear, citing multibillion-euro GDPR fines and a wave of unlawful-recording lawsuits. Every cloud notetaker asks you to trust a vendor with unguarded conversation. Meetily's answer is to never upload it at all.

RelatedStrix Setup: Run an AI Penetration Tester on Your Code

How do you install Meetily on Windows and macOS?

Windows: grab meetily_0.4.0_x64-setup.exe (about 41 MB) from the GitHub Releases page and run it. Windows SmartScreen sometimes flags young open source installers; once you have confirmed the download came from the official Releases page, More info, then Run anyway, is the standard path. macOS: download meetily_0.4.0_aarch64.dmg, open it, and drag Meetily into Applications. Linux has no packaged build yet, so you build from source:

# Linux: build from source (needs Rust, Node and pnpm)
git clone https://github.com/Zackriya-Solutions/meeting-minutes
cd meeting-minutes/frontend
pnpm install
./build-gpu.sh

On first launch, open settings and pick a transcription model. Whisper variants trade speed for accuracy by size, and Parakeet is the project's fast option. The model downloads once, then transcription runs fully offline. GPU acceleration is baked in: Metal plus CoreML on Apple Silicon, CUDA for Nvidia and Vulkan for AMD and Intel on Windows and Linux.

How do you get AI summaries without the cloud?

Transcripts are only half the product; the summaries come from a language model, and the recommended provider is Ollama running on the same machine. Install it from ollama.com, then pull a small model:

# pull a compact local model for summaries
ollama pull llama3.1:8b

The desktop installer keeps Ollama running in the background, so all that remains is opening Meetily's settings, choosing Ollama as the summary provider, and selecting your model. An 8B model wants roughly 8 GB of free memory; on a thin laptop, pick something smaller. If you are willing to trade some privacy for quality on the summary step only, Meetily also speaks to Claude, Groq, OpenRouter, or any OpenAI-compatible endpoint, including one inside your own company network. The transcript itself still never leaves the device.

How does it compare with Otter and Fireflies?

The cloud tools still win on team features: shared workspaces, search across an organization's meeting history, CRM integrations. Meetily wins everywhere data control is the requirement rather than a preference, which describes legal, healthcare, finance, defense and every startup that signed an NDA this week.

RelatedSet Up OpenAI Codex Inside Claude Code

What are the gotchas before you rely on it?

Four things worth knowing before it becomes your system of record. First, summary quality tracks the model you feed it: an 8B model produces serviceable minutes, not magic, and low-memory machines need smaller models still. Second, speaker identification is not in the free app today: the repository blurb advertises diarization, but as of v0.4.0 the README lists speaker identification as a coming-soon feature of the paid PRO edition, so do not adopt the community edition expecting per-speaker transcripts. Third, Linux users get a source build, not an installer, and need Rust, Node and pnpm on hand. Fourth, local processing does not exempt you from consent: recording laws apply to the capture itself, so keep telling participants they are being recorded. None of these are dealbreakers; all of them are the difference between a happy install and a surprised one.

Our take

The local AI stack quietly crossed the good-enough line for meeting notes, and Meetily's real achievement is packaging rather than research: whisper.cpp-lineage transcription, Parakeet for speed, and Ollama summaries wired together behind one installer a non-engineer can run. Meeting audio is the sharpest possible test for the privacy argument because it is the one data stream containing everyone's unguarded conversation, which is exactly why this repo found 20,000 stars. The thing to watch is the open-core split. PRO funding development is healthy; the free edition stagnating while accuracy upgrades go paid would sour the deal that earned those stars. Today, this is the easiest private notetaker to stand up, and the price argument ends the debate.

Original analysis by GenZTech. Tool documentation: Zackriya-Solutions/meetily on GitHub.