Offline speech to text on Mac
See how local Whisper and Parakeet models keep transcription on your Apple Silicon Mac after the one-time download.
Free · Source on GitHub · macOS
Voice to text for Mac.
Speak. It types. Anywhere.
Press Option plus Space⌥Space to dictate into any app. Runs locally on Apple Silicon — free, no account required, and source available on GitHub.
Speak in Notes, Slack, Mail, Notion, ChatGPT, a terminal, or any other text field. Review the result before it types, or turn on instant paste.
Trust, privacy, and local transcription
- Signed & notarized
- No app telemetry
- Auditable source on GitHub
WhisperKit and FluidAudio run speech models locally on the Apple Neural Engine — audio stays on your Mac in local mode.Read the setup guide
See it work
From voice to a finished prompt in 20 seconds.
This is the real app dictating directly into a coding workspace — no copy and paste, no window switch, and no staged interface.
Three steps
Start dictating in any Mac app in three steps.
- 01
Press Option plus Space⌥Space
A floating HUD with a live waveform appears — from any app, no window switch. Esc cancels anytime. Prefer push-to-talk? Switch to hold-to-record in Settings.
- 02
Speak naturally
Full sentences, paragraphs, casual or technical. Punctuation is inferred; with a streaming model the words appear live as you speak.
- 03
Press again — review, paste
The transcript pops up for a quick edit — hit Return and it lands at the cursor of the focused app. Turn review off and it pastes instantly.
Choose your workflow
Where will voice to text save you time?
Dictate everyday writing, longer AI prompts, or entire meetings. Pick a workflow to see how VoiceToText fits the Mac apps you already use.
Writing
Write emails, notes, and messages without breaking your flow.
Put the cursor where you want the text, speak naturally, review the result, and press Return. VoiceToText handles punctuation and types into the app already in front of you.
Quick update — I finished the proposal and shared it with the team. I’ll send the final version before lunch tomorrow.
- Notes
- Slack
- Notion
Mac speech-to-text guides
Pick the voice workflow you want to improve.
Practical guides for private dictation, faster coding, and meeting transcription on Mac.
Voice to text for coding
Dictate prompts, implementation notes, and terminal commands into Cursor, Claude Code, ChatGPT, and other coding tools.
Record and transcribe meetings
Capture your microphone and system audio together, then transcribe calls locally by default—without a bot in the meeting.
What you get
Private dictation, designed for the Mac.
The essentials stay simple. Local transcription is the default; cloud models and AI actions are optional when you explicitly add your own provider key.
Your audio stays local in local mode
Parakeet and Whisper run on-device, so those recordings are not sent to a speech provider. There is no VoiceToText account or app telemetry; model downloads, update checks, optional cloud engines, and AI actions use the network.
Tap to talk — or hold
Press Option plus Space⌥Space to start, again to stop — or switch to hold-to-record push-to-talk. Esc cancels. Rebind to any shortcut, even Right Control alone.
Native Mac speed
Pure SwiftUI with no Electron overhead — quick to launch, light in Activity Monitor, and built for Apple Silicon.
Works in every app that accepts text
Slack, Mail, Notes, browser address bars, terminals, code editors — if macOS puts a cursor there, VoiceToText types into it.
FAQ
What you need to know before installing.
Eight direct answers about privacy, setup, compatibility, and everyday use.
Is VoiceToText free?
Yes — completely free, forever. VoiceToText has no paid tiers, accounts, or in-app purchases, and its source is available on GitHub for inspection. Download it free.
Does it work offline? Is my voice data sent anywhere?
Yes, transcription works fully offline by default. Local models (Whisper, Parakeet) run on the Apple Neural Engine — your audio never leaves your Mac. The app's only network traffic is one-time model downloads and a daily update check against GitHub Releases. Cloud models are strictly opt-in: audio is sent directly to the provider (OpenAI or ElevenLabs) under your own API key only when you explicitly select a cloud engine, and AI transcript actions use your OpenAI key the same way. VoiceToText itself never receives your audio.
How do I use voice to text on my Mac?
Install from the DMG, grant Microphone and Accessibility permissions, then press Option plus Space⌥Spacein any app, speak, and press it again. The transcript pops up for a quick review — hit Return and it’s pasted at the cursor. Prefer zero friction? Turn review off in Settings and text lands instantly. Prefer hold-to-talk? Switch the shortcut to hold-to-record. Works identically on MacBook Air, MacBook Pro, iMac, Mac mini, and Mac Studio.
Why does it need Accessibility permission?
Accessibility permission is how macOS lets one app type into another. VoiceToText uses it to paste transcribed text at the cursor, register the global shortcut, and catch Esc so you can cancel a recording — it does not log your keystrokes, read your screen, or access any other app’s data. (Only the optional Right Control shortcut needs the separate Input Monitoring permission.) The source code is public if you want to verify exactly how the permissions are used.
What are the system requirements?
The current builds require macOS 15.0 or later and an Apple Silicon Mac (M1 or newer). Intel Macs are not supported because the local models rely on the Apple Neural Engine. Cloud models work on any supported Mac with an internet connection.
How accurate is it? Which models does it use?
You choose from six local models: Parakeet TDT v3 is the fastest on-device option, Whisper Large v3 is the most accurate offline option, and Whisper Large v3 Turbo is the best balance — plus Small, Base, and Tiny for older or space-constrained Macs. For the highest accuracy across accents and technical jargon, bring your own API key and switch to GPT Transcribe, GPT-4o Transcribe, or GPT-4o Mini Transcribe (OpenAI), or Scribe v2 Realtime (ElevenLabs) in Settings → Models.
What apps does VoiceToText work in?
Any Mac app with a text field. Apple Notes, Notion, Obsidian, Bear, Pages, Google Docs, Microsoft Word, Slack, Messages, Mail, Gmail, Outlook, WhatsApp, Discord, Safari and Chrome address bars, ChatGPT, Claude.ai — if macOS puts a cursor there, your voice types into it. No app-specific setup.
Do you collect any usage data or telemetry?
The app collects nothing: no accounts, no analytics, no first-party servers. With a local model your audio generates zero network traffic — the only connections the app makes are model downloads and an update check against GitHub Releases. If you opt in to a cloud model, audio goes directly from your Mac to the provider — VoiceToText is never in that path. The repo is public; verify it yourself or watch traffic with Little Snitch. (This website uses Google Analytics to count visits; the app itself contains no analytics code.)
Ready to dictate
Download VoiceToText — voice to text for macOS.
One DMG. Drag to Applications. Grant Microphone and Accessibility. Press Option plus Space⌥Space and speak.
- 1
Open the DMG
Drag VoiceToText to
/Applications. Takes 5 seconds. - 2
Launch the app
The default local model (Parakeet TDT v3) downloads itself — no setup, no account.
- 3
Grant Microphone and Accessibility
One-time prompt — mic hears you, accessibility types into whatever app you’re in. Revoke anytime in System Settings.
- 4
Press Option plus Space⌥Space, speak, press again.
Review the transcript, hit Return — it lands at the cursor. Prefer hold-to-talk? One switch in Settings.
Why two permissions? Microphone lets the app hear you. Accessibility lets it type into whatever app you’re in. Both stay on-device. Revoke anytime in System Settings.
Requirements: macOS 15.0 or later · Apple Silicon (M1 or newer). Updates are built in — the app checks GitHub Releases and installs new versions in place.