Speech to text on Mac
A speech to text Mac app that types into any app.
Speech to text on a Mac starts with what Apple ships: Dictation, built into macOS. It is free, it is good, and for many people it is enough. VocalFlow is a speech to text Mac app for people who dictate all day and want a different shape of tool: hold one key, talk, let go, and clean text is pasted wherever the cursor is.
It is a free, open-source dictation app for Mac (and Windows), built on Deepgram’s Nova-3 speech recognition, with optional spelling and grammar clean-up by a language model you choose. Below: how it compares with built-in Dictation, and how to set it up.
macOS 13 Ventura or later · Apple silicon and Intel · Hold-to-talk hotkey · Free · MIT license
Speaking is about three times faster than typing.
In a Stanford study, people entering short messages by speech were 2.93 times faster than on a phone keyboard, and desktop typists in a separate study of 168,000 people averaged about the same speed as that phone keyboard. Speech also left slightly more errors in the final text (1.30% against 0.79%), which is the whole case for Mac speech to text with a clean-up pass: speak fast, then let the grammar step tidy it.
- Speaking, English
- 153 WPM
- Typing on a phone keyboard, English
- 52 WPM
- Typing on a desktop keyboard, average
- 51.6 WPM
Speech and phone rows: Ruan et al., IMWUT 2017, short messages on an iPhone 6 Plus under lab conditions. Desktop row: mean of 168,000 volunteers, Dhakal et al., CHI 2018 (doi.org/10.1145/3173574.3174220). · Source: Ruan, Wobbrock, Liou, Ng & Landay (arXiv:1608.07323)
Built-in Dictation or a dictation app for Mac?
macOS Dictation is a toggle: start it, talk, stop it. VocalFlow is push-to-talk: it records only while the key is held. That one difference decides which feels natural to you.
| macOS Dictation | VocalFlow | |
|---|---|---|
| How you start | Microphone key, a shortcut such as pressing Fn twice, or Edit > Start Dictation | Hold a hotkey: Right or Left Option, Right or Left Command, or Fn |
| How it stops | You stop it, or it stops after 30 seconds without speech | When you let go of the key |
| Where speech is processed | On your Mac for supported languages; Keyboard settings show whether it is | Streamed to Deepgram with your own API key. No VocalLabs server |
| Punctuation and clean-up | Automatic punctuation in supported languages | Deepgram punctuation, plus optional spelling and grammar fixes by Groq or OpenRouter |
| Languages | Not every language or region; Apple keeps a list | 30+ through Deepgram, including Hindi, Tamil and Marathi |
| Cost | Free, built in | Free app. Deepgram bills per minute after a $200 credit |
| Also on Windows | No | Yes, Windows 10 and 11 |
macOS Dictation details from Apple’s Mac User Guide, as of Oct 2026. · Source: Apple: Dictate messages and documents on Mac
Teach it the words you actually use.
General speech recognition stumbles on product names, colleagues’ names and jargon. VocalFlow keeps a list of the words you care about and gets better at them as you correct it.
- Focus Words
- Pin the exact spelling of names and jargon. With Nova-3 in English they are passed to Deepgram as key terms, so recognition leans toward them.
- Text expansions
- Save a trigger : replacement pair. Say “my email” and the full address is typed.
- Learns from your fixes
- Type over a word VocalFlow just dictated, within about two minutes, and the corrected spelling is saved as a Focus Word.
- Custom instructions
- One line added to the clean-up prompt every time, such as “use Markdown lists” or “always formal English”.
Voice typing on Mac, set up in four steps.
- 01
Install VocalFlow
Download the .pkg from the VocalFlow page and run it, or drag VocalFlow.app from the .zip into Applications. If you used the .zip, run the one-line Terminal command on the download page to clear the quarantine flag.
- 02
Grant two permissions
Microphone, so it can hear you, and Accessibility, so the hotkey works in every app and the text can be pasted. macOS asks the first time you open it.
- 03
Paste in your Deepgram key
Menu bar icon, Settings, paste the key, Save & Verify. Pick a model and a language. Add a Groq or OpenRouter key too if you want spelling and grammar clean-up.
- 04
Pick a hotkey and speak
Choose Right or Left Option, Right or Left Command, or Fn. Click into any text field, hold the key, talk, and let go.
Where Mac users dictate.
Text arrives through the clipboard and a simulated Cmd+V, so it works in any field that accepts a paste.
- 01Email
Mail and Gmail
Replies and first drafts, said out loud and tidied by the grammar pass.
- 02Chat
Slack
Status updates and long replies without breaking your train of thought.
- 03Notes
Notion, Obsidian and Notes
Meeting notes, outlines and journal entries.
- 04Code
Cursor and VS Code
Prompts, PR descriptions and commit messages, spoken instead of typed.
- 05AI
ChatGPT and Claude
Long prompts in a browser tab or a desktop app.
- 06Shell
Terminal
Anywhere that accepts a paste, including a shell prompt.
