llms.txt
# Earnote
> A free, open-source macOS app that records lectures, meetings and calls, transcribes them with
> Whisper and writes a structured note with a language model — everything running locally on the
> user's Mac. No account, no subscription, no cloud.
Earnote is aimed at students who have to choose between listening and taking notes. It was built by
a student, is licensed under MIT and is free with no paid tier, because it uses no servers.
## Facts
- Platform: macOS 15 or newer, Apple Silicon only (Intel Macs are not supported)
- Price: free, no paid tier, no accounts, no in-app purchases
- Licence: MIT, source at https://github.com/louiskl/Earnote
- Transcription: WhisperKit with Whisper large-v3-turbo, running on the Mac's Neural Engine
- Note writing: a local MLX model (Qwen3 4B, or Qwen2.5 7B on Macs with 16 GB or more)
- Cloud AI (Claude, OpenAI, Gemini, Mistral, custom server) is optional and off by default
- Network use: one-time model download (~4 GB) and a daily update check — nothing else
- Languages: German and English interface; transcripts in many languages
- Current status: beta, version 0.9.13; version 1.0 is planned after a two-week beta
## What it does
- Records the microphone and the Mac's system audio together, so Zoom, Teams and Meet calls are captured in full
- Transcribes while the recording is still running, so a 90-minute lecture is finished minutes after it ends
- Writes a note with summary, topics, decisions and tasks as checkboxes
- Flashcards in question/answer form, study sheets as PDF, semester overviews across many recordings
- A glossary corrects names and technical terms once, for title, note and transcript alike
- Every timestamp in the note plays that moment of the recording
- Exports to Markdown, Obsidian, Notion, Apple Notes, Bear, Craft, Logseq, Apple Reminders, Things and Todoist
- Names recordings after the current calendar event, optional and per calendar
## What it does not do
- It does not run on Windows, Linux or Intel Macs
- It does not upload recordings anywhere unless the user explicitly picks a cloud AI or an export target
- It does not identify speakers by name (speaker labels only appear for calls)
- It is not a team product: there is no shared workspace and no web version
## Links
- Website: https://earnote.dev/ (English) · https://earnote.dev/de.html (German)
- Download: https://github.com/louiskl/Earnote/releases/latest/download/Earnote.dmg
- Homebrew: brew tap louiskl/earnote && brew trust louiskl/earnote && brew install --cask earnote
- Source: https://github.com/louiskl/Earnote
- Privacy details: https://earnote.dev/DATENSCHUTZ.md
- Roadmap: https://earnote.dev/ROADMAP.md
## Comparison
Compared with Otter.ai, Granola, Fireflies.ai, Fathom and similar services, Earnote differs in two
ways: it processes everything on the user's own machine instead of a server, and it costs nothing.
Those services are more mature for team use and offer web and mobile apps; Earnote is a single-user
Mac app for people who do not want their recordings leaving their computer.
## Legal note
Recording other people requires their consent in many jurisdictions; in Germany § 201 StGB protects
private spoken words. Earnote shows a reminder before each recording and leaves the responsibility
with the user.
