VoiceToText View on GitHub

Free · Open source · macOS

Free Offline Dictation for Mac.

Push-to-talk dictation for Mac. 100% local. Free forever.

Hold Option plus Space, speak, release — words are typed into Claude Code, Cursor, Slack, or any focused app. On the Apple Neural Engine. Offline.

Download for Mac View on GitHub

Free · macOS 14+ · Apple Silicon

No account. No subscription. No cloud.

Open source on GitHub

Powered by OpenAI Whisper (via WhisperKit) and Parakeet (via FluidAudio), on the Apple Neural Engine. Read the code

See it work

See VoiceToText dictate into any Mac app.

Hold the hotkey. Speak a full sentence. Release. Your words appear at the cursor — in a code editor, a terminal, a Slack thread, a note.

VoiceToText demo: user holds Option plus Space and speaks a prompt; transcribed text is typed into Claude Code in real time on macOS. Menu-bar icon shows recording state.
Dictating a prompt into Claude Code. Nothing leaves the Mac.

How push-to-talk dictation works on Mac.

Three keys. One loop. No menus, no windows, no copy-paste.

  1. 1

    Hold Option plus Space

    Press and hold the global hotkey from any app. A waveform appears in the menu bar.

  2. 2

    Speak naturally

    Say a word, a sentence, or a full paragraph. Transcription runs on-device on the Apple Neural Engine.

  3. 3

    Release — words are typed

    Let go of the keys. Your text is typed straight into the focused app at the cursor.

Rebind the hotkey in Settings. Works in any text field.

Why VoiceToText?

A native Mac app that keeps your voice on your Mac. Free, open, fast, and designed to live at your cursor.

  • Offline and private

    Audio is transcribed locally on the Apple Neural Engine. No cloud, no telemetry, no network calls.

    Read the source

  • True push-to-talk

    Hold Option plus Space to dictate. Release to stop. No toggles, no cutoffs mid-sentence, no rebound.

  • Apple Silicon native

    A SwiftUI menu-bar app that rides the Neural Engine. No Electron, low battery drain, fast cold start.

  • Built for AI agents

    Dictate prompts into Claude Code, Cursor, Codex, Copilot Chat, or ChatGPT at thinking speed.

  • Two engines, your pick

    WhisperKit (OpenAI Whisper) for accuracy. FluidAudio (Parakeet) for speed. Switch in Settings anytime.

  • Types into any app

    Slack, Mail, Notes, browsers, terminals, code editors — anywhere macOS shows a text cursor, words land there.

Voice prompting

Voice dictation for Claude Code, Cursor, and AI coding agents.

Think in paragraphs. Speak them into the editor. Prompts stay on your Mac — no cloud STT sitting between you and the agent.

Dictate the whole prompt: the file you want changed, the error, the constraint, the style you're aiming for. VoiceToText types it at the cursor in Claude Code, Codex CLI, Cursor, Copilot Chat, ChatGPT, Warp, or a plain terminal. Proprietary code never leaves the Mac.

  • Claude Code
  • Cursor
  • Codex CLI
  • Copilot Chat
  • ChatGPT
  • Warp
  • Terminal
  • VS Code
▌ refactor this function so it streams
▌ tokens instead of buffering, keep the
▌ type signature, add a test with fakes
Spoken into Claude Code in one push. No copy-paste, no separate window.
Download for Mac

A free, open-source alternative to Wispr Flow, Superwhisper, and MacWhisper.

Same core idea — dictate into any app. Different terms: no subscription, no account, no cloud, and source you can audit.

Feature
VoiceToText
Wispr Flow
Superwhisper
MacWhisper
Apple Dictation
Price
Free forever
$12–15 / mo
$8.49 / mo · paid lifetime2
$69–80 lifetime3
Free (built-in)
Open source
Yes — on GitHub
No
No
No
No
Runs 100% offline
On-device
Cloud only
On-device
On-device
On-device (some langs)
No account required
None required
Account required
Account required
No account
None required
Push-to-talk (hold key)
Yes. Hold Option plus Space
Customizable hold
Customizable hold
Pro (Global mode)
Toggle only
Works in any text field
System-wide
System-wide
System-wide
Pro only
System-wide
Apple Neural Engine acceleration
Native
— Not advertised1
Yes
Yes
Yes
Native macOS app (no Electron)
SwiftUI native
Cross-platform shell
Native
Native
Native
Multiple speech engines
WhisperKit + Parakeet
Proprietary only
Whisper variants
Whisper variants (Pro)
Apple only
Tuned for AI coding agents
Yes — headline use case
Works in apps
Messaged
Works in apps
No

VoiceToText

Open source
Price
Free forever
Open source
Yes — on GitHub
Runs offline
On-device
No account
None required
Push-to-talk
Yes. Hold Option plus Space
Any text field
System-wide
Neural Engine
Native
No Electron
SwiftUI native
Engines
WhisperKit + Parakeet
AI coding agents
Headline use case

Wispr Flow

Price
$12–15 / mo
Open source
No
Runs offline
Cloud only
No account
Account required
Push-to-talk
Customizable hold
Any text field
System-wide
Neural Engine
— Not advertised1
No Electron
Cross-platform shell
Engines
Proprietary only
AI coding agents
Works in apps

Superwhisper

Price
$8.49 / mo · paid lifetime2
Open source
No
Runs offline
On-device
No account
Account required
Push-to-talk
Customizable hold
Any text field
System-wide
Neural Engine
Yes
No Electron
Native
Engines
Whisper variants
AI coding agents
Messaged

MacWhisper

Price
$69–80 lifetime3
Open source
No
Runs offline
On-device
No account
No account
Push-to-talk
Pro (Global mode)
Any text field
Pro only
Neural Engine
Yes
No Electron
Native
Engines
Whisper variants (Pro)
AI coding agents
Works in apps

Apple Dictation

Price
Free (built-in)
Open source
No
Runs offline
On-device (some langs)
No account
None required
Push-to-talk
Toggle only
Any text field
System-wide
Neural Engine
Yes
No Electron
Native
Engines
Apple only
AI coding agents
No

Comparison table footnotes

  1. 1 Wispr Flow does not advertise Apple Neural Engine acceleration; treat as unverified.
  2. 2 Superwhisper lifetime pricing was in flux at research — verify on the vendor site.
  3. 3 MacWhisper Pro price varies across sources ($69, $79.99, €59) — verify on the Gumroad product page.

Read the source · See the license

Frequently asked questions.

Answers to what developers, writers, and privacy-conscious users ask before installing.

Is VoiceToText really free?
Yes. VoiceToText is 100% free and open source, released under an OSI-approved license. There are no paid tiers, no accounts, no telemetry, and no in-app purchases.
Does VoiceToText work offline?
Yes. All speech recognition runs on-device on the Apple Neural Engine. Your audio never leaves your Mac, and no internet connection is required after installation.
How accurate is it?
Accuracy comes from the models: OpenAI's Whisper (via WhisperKit) and NVIDIA's Parakeet (via FluidAudio) — the same state-of-the-art models behind popular paid apps. Pick the engine that fits your accuracy-versus-speed preference.
Which speech recognition models does VoiceToText use?
VoiceToText ships with two engines: WhisperKit (OpenAI's Whisper) for maximum accuracy, and FluidAudio (NVIDIA's Parakeet) for maximum speed. Both run locally on the Apple Neural Engine.
Can I use VoiceToText to dictate prompts into Claude Code, Cursor, or other AI coding tools?
Yes. VoiceToText types into whatever app has focus, including Claude Code, Codex CLI, Cursor, Copilot Chat, ChatGPT, and any terminal or code editor. It is purpose-built for voice prompting AI coding agents at natural speaking speed.
What Macs does VoiceToText support?
VoiceToText requires macOS 14 (Sonoma) or later and an Apple Silicon Mac (M1 or newer). Intel Macs are not supported because there is no Apple Neural Engine.
How is VoiceToText different from Apple Dictation or Wispr Flow?
Apple Dictation is toggle-style and tied to Apple's models. Wispr Flow is a paid subscription that processes audio in the cloud. VoiceToText is free, open source, held push-to-talk, and 100% on-device.
Do you collect any data?
No. No accounts, no telemetry, no network calls. The repo is public — inspect the source or watch the network with Little Snitch to verify.

Ready to dictate

Download VoiceToText for macOS.

One DMG. Drag to Applications. Grant Microphone and Accessibility. Hold Option plus Space and speak.

Download for Mac Browse Releases

Free · Open source · macOS 14+ · Apple Silicon

  1. 1

    Open the DMG

    and drag VoiceToText to /Applications.

  2. 2

    Launch the app

    It lives in the menu bar.

  3. 3

    Grant Microphone and Accessibility

    when prompted — mic captures audio, accessibility types into the focused app.

  4. 4

    Hold Option plus Space, speak, release.

    Your words are typed at the cursor.

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 14 Sonoma or later · Apple Silicon (M1 or newer).