Echo

What does Echo do?

Echo describes itself as "Record, transcribe and summarise your meetings on your own Mac." A browser extension captures the tab audio from a web meeting; a desktop app transcribes with Whisper Large V3 Turbo, writes a summary with a local Ollama model, and keeps everything on the Mac. There is no bot that joins the meeting.

The README's "no bot" line matters. Otter.ai and Fireflies.ai send a bot into a Google Meet call, which joins the meeting room and becomes a participant everyone sees. Echo's approach keeps the user as the only participant, with the browser extension as the capture point, which the README calls out as a privacy choice. Speaker separation is done with PyAnnote and WavLM after capture.

Key facts

• Licence: MIT.
• Platform: Apple Silicon Mac on macOS 14 or later.
• Meeting platforms: Google Meet, Zoom (web), Microsoft Teams (web), through a browser extension for Brave or Chrome.
• Also: in-person recording via microphone, and uploading an existing audio file.
• Transcription: Whisper Large V3 Turbo, running on the Mac, offline after the one-time model download. Whisper Large V3 is also supported.
• Speaker labels: PyAnnote and WavLM.
• Summaries: a local Ollama model, with action items and decisions.
• Storage: encrypted on the Mac with SQLCipher. Keys are stored in the macOS Keychain.
• Calendar: a read-only Google Calendar connection.
• Build requirements: Xcode command-line tools, Rust, Node.js 22.18 or later, pnpm.
• Install: clone the repo, install the browser extension, build and run the desktop app. Served locally at http://localhost:3000.

What does it replace, and where does it fall short?

Echo replaces Otter.ai and Fireflies.ai for Apple Silicon Mac users who want to keep meeting audio on the Mac and avoid having a bot join the call.

Where it falls short: Apple Silicon Mac only on macOS 14 or later. Intel Macs, Windows and Linux are not supported. The browser extension works in Brave or Chrome; Safari, Firefox, Edge and native meeting apps (the Zoom and Teams desktop apps) are not listed. Install is build-from-source and needs Xcode command-line tools, Rust, Node 22 and pnpm, which is more setup than a drag-and-drop app. There is no cross-team sharing feature in the stock build; the data stays on one Mac.

How does Echo run?

Install the prerequisites: Xcode command-line tools, Rust, Node.js 22.18 or later, and pnpm. Clone the repository and install the browser extension in Brave or Chrome. Build the desktop app, which serves the UI at http://localhost:3000 on the Mac.

To record a meeting, open the web meeting in Brave or Chrome and use the extension. The audio is captured, transcribed on the Mac with Whisper Large V3 Turbo, and summarised with the local Ollama model. Everything is encrypted at rest with SQLCipher and keys are stored in the macOS Keychain.

Who is Echo for?

An Apple Silicon Mac user who runs Ollama already and wants a local, build-it-yourself meeting notes tool that never calls a cloud service. A team with mixed hardware or shared notes is better served by Scribe or Meetily. A reader who wants a pre-built disk image rather than source code should look at those too.

What limits does the README state?

• Apple Silicon Mac only, macOS 14 or later.
• Browser extension in Brave or Chrome only; other browsers are not supported.
• No bot joins the call; capture is through the extension on the user's browser.
• Build-from-source install: Xcode command-line tools, Rust, Node 22.18 or later, and pnpm are required.
• Google Calendar is read-only.
• The README does not list measured accuracy or latency.

Questions people ask

Is Echo open source?

Yes. The LICENSE file at the repository root is MIT, an OSI-approved licence. Whisper Large V3 Turbo is MIT by OpenAI. The PyAnnote and WavLM speaker-labelling models have their own licences on Hugging Face, which the user accepts on first download. Ollama is MIT.

Does Echo work offline?

Yes, in its default configuration. The README says "No bot joins your call, there is no account, and your audio never leaves your laptop." Whisper runs on the Mac through the Rust and Swift desktop app, and the Ollama summary model runs locally too. The only external calls are Google Calendar for the read-only meeting list, which can be skipped.

Can Echo be used commercially?

Yes for the code, which is MIT. Whisper's weights are MIT. The PyAnnote and WavLM speaker models have their own licences on Hugging Face, and a commercial user should read them before shipping. Ollama is MIT. Echo itself adds no extra restriction.

How does Echo compare with Otter.ai and Fireflies.ai?

Otter.ai and Fireflies.ai send a bot into the meeting room, keep audio on their servers, and bill a monthly fee per seat. Echo captures audio through a browser extension on the user's own machine, transcribes and summarises on the Mac, and bills nothing. The trade is that Echo is Apple Silicon only, browser-based capture only, and build-from-source. The open-source alternatives to Otter.ai and Fireflies.ai pages set it beside Scribe and Meetily.

Does Echo need Ollama?

Yes, to produce summaries, decisions and action items. The README says summarisation runs through a local Ollama model, so Ollama needs to be installed and a model pulled (the README does not fix a specific one). Transcription with Whisper works without Ollama, so a reader who only wants transcripts can skip the summary step.

Sources

• Echo README and LICENSE: github.com/samarthmn/echo-voice-web, read 2026-11-09.