Free · On-device · macOS
Dictation lives in one subscription. Meeting notes live in another. Better Voice does both jobs in a single Mac app. Hold a key to dictate anywhere, record a meeting and get a speaker-labeled summary. Then it files the result in Apple Notes, where you already keep everything. It all runs on your Mac, so your audio never leaves it.
Free and MIT-licensed. One first-run download puts the speech model on your Mac. Here's what setup actually involves.
These are normally two separate purchases from two separate companies, each with its own app to open, its own bill, and its own silo your words end up in. Better Voice does both, and hands the result to Apple Notes.
Job one
Hold your hotkey, talk, let go. The text appears at your cursor: email, Slack, your editor, a form in the browser. There's no window to switch to and nothing to paste. Filler words come out by a fixed word list and your own spellings go in by exact word-match, so nothing rewrites what you actually said.
Instead of a dictation subscription
Job two
Record the call straight off your Mac, drop in a recording you already have, or paste a transcript. Better Voice transcribes it, works out who spoke when, has you name the voices once, and writes a summary with a real title. Name someone once and it recognises them in later meetings.
Instead of an AI notetaker
Meetings
Thursday, June 18, 2026 at 9:03 AM · 27 min
Priya, Sam, and Jordan reviewed the Q3 roadmap. The API migration is unblocked now that the auth refactor landed, and the team agreed to ship the beta by Friday.
Key points
Action items
Every meeting shows up as a titled note in Apple Notes, ready to search, edit, and share.
Two products, two bills, and your words split across both of them, plus whichever cloud each vendor keeps your recordings in.
Not a library inside another app you have to remember to open. Every meeting becomes a titled note with summary and transcript, in folders you pick once. Searchable in Spotlight, editable on your phone, and already synced to your iPad by the time you shut your laptop.
Once the speech model is down, dictation and transcription need no connection at all. They're running on your Mac, not on someone's API. Only the Apple Notes sync at the end wants the internet, and it waits.
Better Voice records the audio on your machine instead of attending as a guest. Nobody else in the meeting has to accept a third-party notetaker, and there's no participant list entry explaining what you're using.
Name a voice once and Better Voice suggests that name next time it hears them. Naming gets faster every meeting instead of starting over.
No subscription, no account, and no tier that holds back the useful half. The source is on GitHub.
Signed with a Developer ID and notarized by Apple, so it installs by drag-and-drop with no Gatekeeper workaround. Updates arrive in-app, and your permissions carry across them.
Two things are worth knowing before you download. Both are real, and neither is buried in a settings pane.
Required · automatic · once
On first launch Better Voice downloads NVIDIA's Parakeet speech model: about 470 MB, a few minutes, once. That download is exactly why your audio never has to leave your Mac afterwards. The menu bar shows progress, and dictation switches on the moment it lands.
Optional · Ollama for long meetings
Transcription needs no setup at all. Writing the summary does need a language model. Apple's on-device one is zero setup but has a short context window, so a long meeting gets chunked. For hour-long calls, point Better Voice at Ollama or any OpenAI-compatible server you run. Either way it stays on your machine.
AI notetakers upload your meetings to their servers to transcribe and summarize them. Better Voice does all of that on your Mac. The only thing that ever leaves is the finished note and only to your own iCloud, the same way every other Apple Note syncs. We run no servers.
BetterVoice2.app into your Applications folder.