Meetily

Run Whisper or Parakeet transcription on the device and keep recordings and transcripts local

The README presents local meeting capture, real-time transcription, and local storage of recordings and transcripts as core Community Edition features. Transcription can use Whisper or Parakeet without sending the audio to an external server.

The Tauri backend is written in Rust and combines audio capture, Whisper, Parakeet through ONNX Runtime, and SQLite. Packaged macOS builds use Metal/CoreML, while the Windows package is documented with Vulkan support for local inference acceleration.

Sources: [2][6]

Choose the summary provider: Ollama for local processing or external LLM services when needed

Meeting summaries can use local Ollama, Claude, Groq, OpenRouter, or an OpenAI-compatible endpoint. This also allows an organization to point Meetily at an internal LLM gateway or compatible endpoint.

Privacy therefore needs to be evaluated per feature. Audio capture and transcription are local-first, while an external summary provider receives the meeting content needed for summarization. Use Ollama or an organization-controlled endpoint when the summary path must remain local.

Sources: [2][4]

For meetings where recordings and transcripts should stay on a managed device rather than a meeting SaaS

Meetily is relevant for client calls, internal engineering meetings, research, legal, healthcare, or other workflows where sending recordings and transcripts to a hosted meeting bot is undesirable. It also fits workflows that prefer capturing microphone and system audio from a desktop application instead of inviting a bot into the meeting platform.

Community Edition and Meetily PRO are separate codebases; do not merge their feature lists

The README says the Community Edition remains free and open source, while Meetily PRO is a separate professional codebase. Custom summary templates, advanced exports, automatic meeting detection, and team-oriented self-hosted deployment are described as PRO capabilities.

Adoption decisions should therefore distinguish features available in this GitHub repository from capabilities marketed for PRO or Enterprise on meetily.ai.

Sources: [2]

The Windows package requires an AVX2-capable x64 CPU, and Linux currently follows a source-build path

The README says the standard Windows installer uses a Vulkan-enabled Whisper build and requires an AVX2-capable x64 CPU. CUDA acceleration requires a source build with a compatible NVIDIA toolchain. Packaged macOS builds use Metal and CoreML.

For Linux, the documented installation path is a source build rather than a packaged installer. Fleet deployment should account for CPU instruction support, GPU backend, and the update/build process in addition to operating-system compatibility.

Sources: [2][6]

Analytics is opt-in, but external LLMs and enabled analytics create outbound traffic

The Privacy Policy describes usage analytics as opt-in and off by default. If enabled, anonymized feature usage, session, and performance data is sent to PostHog; meeting content, transcripts, and recordings are stated to be excluded.

External LLM providers are optional as well. Treat Meetily as fully offline only when provider and analytics settings are configured accordingly.

Sources: [4][2]

MIT-licensed application code, with separate obligations for bundled models and third-party components

Meetily itself is published under the MIT License. The README also acknowledges code from Whisper.cpp, Screenpipe, and transcribe-rs and uses the NVIDIA Parakeet model ecosystem.

For custom builds or internal redistribution, review the licenses and usage terms of the application, model weights, runtimes, and other bundled dependencies rather than treating the application MIT license as covering every artifact.

Sources: [7][2][6]