Open-source Otter.ai alternatives

Meetily is an open-source AI meeting assistant that records, transcribes and condenses meetings entirely on your local machine, so audio and notes never leave your infrastructure. It is aimed at professionals and enterprises that need meeting intelligence without giving up privacy, compliance or control.

Transcription runs locally using Whisper or Parakeet models, and summaries are generated through Ollama with open-source language models, which avoids paying for external APIs. It works offline and supports multiple meeting platforms. The project mentions speaker diarization, and the desktop app is built with Rust for macOS and Windows. Because it is open source, you can self-host and modify it for your own needs.

The community edition is MIT licensed. A paid Meetily PRO tier is offered for users who want better accuracy, richer export options, customizable summary workflows and team features. It suits people who sit in many calls and want searchable notes without sending recordings to a third-party cloud.

Buzz is a desktop application that turns recorded speech into text and translates it, all on the user's own computer. It builds on OpenAI's Whisper speech recognition models, so audio files do not have to be uploaded to a third-party transcription service. The project is written in Python and released under the MIT license.

The app is distributed for Windows, macOS and Linux. Windows users get an installer, Linux users can choose Flatpak, Snap or AppImage packages, and a PyPI package exists for people who prefer a pip-based setup with ffmpeg installed. Nvidia GPU acceleration through CUDA is supported for faster transcription, and recent releases require Apple silicon on Macs.

Because processing happens locally, Buzz suits people who handle sensitive recordings or want to avoid per-minute fees from cloud transcription services. It runs as a local program rather than a hosted service, so there is no account to create.

Pluely is a desktop AI assistant built with Tauri, Rust, and React. It runs as a translucent overlay with two modes, Ask and Listen, plus a dashboard for chats, meetings, files, prompts, and settings. Ask lets you query anything visible on screen, and Listen transcribes a conversation and suggests answers as it goes. It is positioned as an open alternative to Cluely.

The overlay is designed to stay out of screen shares, and no bot joins a call. The project lists interviews, sales calls, standups, lectures, and live debugging as moments where switching to a browser tab would be disruptive. Builds are provided as .dmg for macOS, .msi or .exe for Windows, and .deb, .rpm, or AppImage for Linux, with a free plan and no account needed to start.

An important change is described in the README: version 1 is distributed as signed binaries rather than open source, because the earlier code was repackaged and sold as clones. The input metadata still lists GPL-3.0, so check the repository for current terms before relying on it as open source.

Call.md turns meetings into live agent loops: it records, transcribes and analyzes meetings with AI before, during and after calls. It is an Electron desktop app, built with React and TypeScript, and it works alongside AI agents through MCP and skills.

Installation on macOS 12 and later, on Apple Silicon and Intel, uses a curl command. After launching, you grant microphone and screen-recording permissions, register with a VideoDB API key, and can optionally connect Google Calendar. The README notes that the VideoDB capture SDK ships recording binaries for macOS on both architectures and Windows x64; Windows x64 can be built from source with no hosted installer yet, while Linux builds have app features but recording is not supported and Windows ARM64 is not supported.

Because recording and capture rely on the VideoDB service and an API key, it is not a fully offline tool. The repository metadata does not state a license, so check the project terms before using or redistributing it.