Two ways to talk to your Mac
There are two broad approaches to voice to text on a Mac today.
Transcription writes down what you said, word for word, with punctuation. The dictation built into macOS works this way. What you say is what you get.
Transcription plus cleanup writes down what you said and then edits it the way a careful assistant would: filler words go, sentences get their commas, and the result is shaped for where it is going. Reeve works this way. You hold fn, talk, let go, and the finished text is typed where your cursor is, usually 1.5 to 2.5 seconds after you release the key.
What the built-in dictation does well
Apple's dictation deserves credit. It is free and already on your Mac. On Apple Silicon Macs it processes speech on the device for many languages, so for those languages your voice does not have to leave the computer. It adds punctuation automatically in supported languages and understands spoken commands for things like new lines. For short messages in one language, it does the job.
Where Reeve is different
| Built-in dictation | Reeve | |
|---|---|---|
| Price | Free | $19 once + provider usage (≈ $1.50/month at 50 dictations a day) |
| Where speech is processed | On the device for many languages on Apple Silicon | At the AI provider you choose, with your own key |
| Filler words and false starts | Written as spoken | Removed |
| Formatting per app | Same output everywhere | Rules per app and window title, editable |
| Rewrite marked text by voice | No | Yes |
| Remembers your corrections to names and terms | No | Yes |
| Danish and English in one sentence | One language at a time | Mixed freely, plus many other languages |
| Works offline | Yes, for on-device languages | No |
Output shaped for the app you are in
Before Reeve writes a word, it checks which app and window you are in. In a terminal it writes one line. In Slack it stays casual. In Mail it writes paragraphs. You can change these rules for any app, and a rule can match on the window title too, so a GitHub pull request in the browser can get different treatment from a news site in the same browser.
Edit by voice
Mark a paragraph, hold the key and say what to change: “make this shorter”, “more formal”, “turn this into a bullet list”. Reeve replaces the marked text with the rewritten version. The emails and messages guide has examples.
Names and jargon
Client names, product names and in-house terms are where dictation usually stumbles. When you fix how Reeve wrote a word, it keeps your version for next time.
Long dictations and changing microphones
For long dictations, Reeve sends the audio in slices of about 20 seconds while you are still talking, so five minutes of speech arrives a few seconds after you stop. If you put in AirPods halfway through, it keeps recording on the new microphone. Every result also lands on your clipboard as a backup. More on this in the long dictation guide.
Which one should you use?
Stay with the built-in dictation if you mostly send short messages in one language, you want speech to stay on the device, or you want to spend nothing.
Try Reeve if you write a lot of mail, documents or code-adjacent text, you switch between Danish and English, your work is full of names that dictation gets wrong, or you are tired of cleaning up “um” and “so yeah” after every dictation.
Nothing stops you from using both. Reeve sits on the fn key. If the built-in dictation also uses that key on your Mac, move it to another shortcut in System Settings, under Keyboard.
What you need
- macOS 14 or later on an Apple Silicon Mac.
- Microphone and Accessibility permission for Reeve. Accessibility is what lets it type into other apps.
- A key from a supported AI provider: OpenAI for transcription and clean-up, or ElevenLabs for transcription (add an OpenAI key for the clean-up). See support for setup.
A Windows version is in progress.