Meetily: Privacy-First AI Meeting Assistant With Local Whisper and Parakeet

Cloud meeting assistants send your audio to someone else's GPU. Meetily — ~21.3k GitHub stars, 2.1k forks, MIT license — takes the opposite bet: capture, transcribe, and summarize on your machine, with optional cloud LLMs only if you wire them in.

Latest release: v0.4.0 (June 5, 2026). Stack: Tauri (Rust backend + Next.js frontend), local Whisper or Parakeet for speech-to-text, and Ollama (recommended), Claude, Groq, or OpenRouter for summaries.

If you already run local dictation with Parakeet or self-host models via Ollama, Meetily extends the same privacy posture to meetings — the category where "just use Zoom AI" is often a compliance non-starter.

TL;DR: What People Are Asking

What Meetily Does

Meetily is a single self-contained app — not a SaaS dashboard with a desktop wrapper.

Meeting audio (mic + system)
        │
        v
Local STT ── Whisper.cpp lineage OR Parakeet (GPU)
        │
        v
Live transcript + local storage
        │
        v
Summary LLM ── Ollama (local) / Claude / Groq / OpenRouter / custom endpoint
        │
        v
Meeting notes on your disk

Privacy-first by architecture: recordings, transcription models, and transcripts stay local unless you explicitly route summaries to a cloud API.

Professional audio: simultaneous mic + system capture with ducking and clipping prevention — the unglamorous feature that separates usable meeting tools from demoware.

Key Features in v0.4.0

Local live transcription

• Whisper — proven open-weight STT; Meetily borrows from whisper.cpp ecosystem patterns
• Parakeet — NVIDIA's model; ONNX conversion credited to istupakov; marketed as faster for live capture
• GPU acceleration baked into builds:
  • macOS: Apple Silicon Metal + CoreML
  • Windows/Linux: NVIDIA CUDA, AMD/Intel Vulkan

No toggle maze — acceleration enables when hardware supports it.

Import & Enhance (beta)

Contributed by Jeremi Joslin; improved by Vishnu P S and Mohammed Safvan.

Import existing audio files → generate transcripts, or re-transcribe any recorded meeting with a different model or language — all local. Useful when your live pass used a fast Parakeet model but you want a higher-quality Whisper pass offline.

AI-powered summaries (provider-flexible)

This mirrors the closed vs local AI split: keep the sensitive layer (raw audio + first-pass text) local; optionally use frontier APIs only on redacted excerpts if policy allows.

Custom OpenAI endpoint

Organizations with private LLM gateways can point summaries at internal endpoints — same pattern teams use for building local-first AI workflows.

Installation

Windows

1. Download latest x64-setup.exe from GitHub Releases
2. Run installer

macOS

1. Download meetily_0.4.0_aarch64.dmg
2. Drag to Applications
3. Launch from Applications

Linux (build from source)

git clone https://github.com/Zackriya-Solutions/meetily
cd meetily/frontend
pnpm install
./build-gpu.sh

README also references meeting-minutes as an alternate clone path — the canonical repo is Zackriya-Solutions/meetily. You need Rust and Node.js; see the repo's Building on Linux docs for distro-specific deps.

Architecture (For Developers)

If you're comparing to Screenpipe-style continuous capture tools, Meetily is meeting-scoped — optimized for scheduled calls and explicit recording sessions, not 24/7 screen indexing.

Meetily PRO vs Community Edition

Community Edition promises to stay free and open source forever with local transcription, summaries, and core UX.

Meetily PRO (meetily.ai/pro) is a separate codebase targeting professionals and teams:

LAUNCH20 — 20% off PRO until the next Community Edition release (community thank-you coupon).

Enterprise — 100+ users, managed compliance — separate sales path.

Honest framing: PRO is how Zackriya monetizes while keeping CE open. If you need speaker labels today, check PRO status or plan manual speaker tags in CE transcripts.

Privacy and Compliance Angles

Recent repo work matters for evaluators:

• fix: make analytics opt-in by default — telemetry off unless you enable it
• PRIVACY_POLICY.md in repo — read before enterprise rollout
• Local-first STT — audio never required to hit a vendor cloud for transcription
• Summary provider choice — compliance team can mandate Ollama-only configs

Contrast with Aleph Neuro silent speech research (private input) or FluidVoice (private dictation): Meetily covers the meeting-length, dual-channel audio problem — harder storage and retention rules.

Limitations (Read Before You Rely on It)

Meetily is not a replacement for legal hold, e-discovery, or regulated healthcare transcription without your own validation pass.

Who Should Use Meetily?

Good fit:

• Engineers and PMs who already run local LLMs
• Consultants under NDA who cannot ship client calls to Otter/Fireflies
• EU teams aligning with sovereign AI / data residency narratives
• Open-source contributors who can patch Rust STT pipelines

Poor fit:

• Teams wanting zero-setup cloud sync and shared workspaces out of the box
• Users who will not manage GPU drivers or model downloads
• Organizations needing certified diarization today without PRO evaluation

The Bottom Line

Meetily is the most starred open-source, local-first meeting assistant in mid-2026 — 21.3k stars for a reason. v0.4.0 delivers real-time Whisper/Parakeet transcription, Import & Enhance re-runs, and flexible summary routing with Ollama as the privacy-preserving default.

It inherits DNA from whisper.cpp and Screenpipe, ships as a polished Tauri app, and draws a clear line between free Community Edition and commercial PRO.

Fork it if you need control. Install the dmg/exe if you want notes without sending audio to the cloud. Read the open issues before you bet your compliance story on CE alone.

Related on explainx.ai

• Inflect-Micro-v2 — local TTS under 10M params — on-device speech output sibling to local STT
• FluidVoice: Local macOS Dictation with Parakeet — same local STT philosophy, dictation use case
• Aleph Neuro Silent Speech — private voice input research
• Closed-Source vs Local Open-Source AI — when to keep models on-device
• How to Run Open-Source Models Locally — Ollama setup patterns
• Build a Personal Local AI System — workflow design
• Ollama 0.31 and Faster Local Agents — local LLM runtime updates
• Voicebox: Open Source Voice Studio — adjacent audio AI tooling
• GPT Realtime 2 Voice API — cloud alternative for speech agents
• AI Regulation 2026 — compliance context for meeting data

Source: Zackriya-Solutions/meetily on GitHub (MIT, README and v0.4.0 release as of July 8, 2026).

Star counts, PRO feature availability, and coupon terms reflect public repo and meetily.ai marketing at publication time.