Open source, MIT licensed

Dictate into any app,
with models you choose

Hold a shortcut, speak, and the finished text lands in the field where your cursor was. Run speech models locally on your machine, or connect directly to cloud providers with your own API keys.

Download Source on GitHub

Free, no account or subscription · Mac, plus iPhone, iPad, Linux and Windows betas

Mac demo with narrated captions and fictional sample data (clipboard mode on the CI test runner).MP4 ↗

How it works

Hold a key, speak, and keep working

Just Speak to It stays in the background while another app is active, so dictating does not pull focus away from your window.

01

Trigger from anywhere

Hold fn, double-tap it, or bind a custom global shortcut. Recording starts while the frontmost app keeps focus.

02

Transcribe and apply rules

Stream words live as you speak or transcribe a saved audio file. Personal corrections, pronunciation hints and per-app profiles adjust the model and clean-up prompt for code, chat or email.

03

Insert where you started

The session records the active text field before capture begins and types the result back into it, even if another window came forward mid-sentence. Clipboard output is optional and restores the previous clipboard afterwards.

Privacy and keys

No middleman server

Audio stays on the device when a local model runs, and goes straight to the chosen provider when a cloud key is configured. Just Speak to It runs no relay server and requires no account.

Local models run offline

Mac builds include Apple's on-device speech recognisers, six WhisperKit models on Core ML, and Parakeet for local streaming. Windows and Linux betas run whisper.cpp locally. Clean-up can use Apple Intelligence, downloaded local LLMs or offline rules.

Cloud providers use your keys

Cloud transcription, clean-up and speech synthesis use accounts you control. Keys live in the system credential store, and requests go directly to each provider (or through an optional local Azure proxy on your Mac).

Private history and opt-in telemetry

History stays on the device unless synced through a private iCloud container. Crash diagnostics and optional Mac usage analytics exclude audio, transcripts and keys.

Apple's older Speech Recognition engine can fall back to Apple's servers when an on-device model is unavailable, and enabling iCloud sync or cloud providers sends data off the device. Read the privacy policy for the full breakdown.

Models

Switch models per task or compare them side by side

Pick a fast streaming model for short replies, an offline model on a train, or compare two providers on the same recording with estimated cost in History.

Live dictation

  • Apple (on-device)
  • Parakeet
  • whisper.cpp
  • OpenAI
  • Deepgram
  • AssemblyAI
  • ElevenLabs
  • Soniox
  • Mistral
  • Speechmatics
  • Cartesia
  • Gladia
  • Google
  • Modulate
  • Meta
  • Rev.ai
  • xAI

Streaming models display words while you speak.

Recorded files

  • Apple (on-device)
  • WhisperKit
  • Parakeet
  • whisper.cpp
  • OpenAI
  • Deepgram
  • AssemblyAI
  • ElevenLabs
  • Soniox
  • Groq
  • Mistral
  • Speechmatics
  • Cartesia
  • Gladia
  • Rev.ai
  • Google
  • Modulate
  • Meta
  • xAI
  • OpenRouter

Batch engines transcribe voice memos, interviews and meeting recordings.

Transcript clean-up

  • Apple Intelligence
  • Local LLMs
  • Offline rules
  • OpenAI
  • Claude
  • Gemini
  • Qwen
  • Llama
  • Mistral

An optional pass fixes punctuation and filler words, or follows a custom prompt through local models or OpenRouter.

Voice output

  • System voices
  • ElevenLabs
  • OpenAI
  • Azure
  • Deepgram
  • Soniox
  • Cartesia
  • Groq
  • Google
  • Speechmatics
  • xAI

Built-in system voices and cloud providers read selected text or clipboard contents aloud.

Model availability varies by platform and OS version. Full provider list → Request a model →

iPhone & iPad

Hands-free capture from the Lock Screen

Start recording from a hardware press or Control Centre, check progress in the Dynamic Island, and send finished text to the clipboard or your Mac.

  • Action Button, Control Centre and SiriStart or toggle capture without opening the app. The Dictate Shortcut action stops automatically after a pause and passes the text to the next step.
  • Live Activity with a Stop controlThe Dynamic Island shows recording state, word count and recent words, with a button that ends the session directly.
  • Configurable outputSet once whether hardware triggers copy raw text, copy and polish, or save quietly to History.
  • Send to Mac via iCloudPhone captures arrive in Mac History through your private iCloud container, with Handoff support and optional paste.
Simulator tour with fixed sample text. Full-size screenshot ↗

Automation

Scriptable from the terminal, Shortcuts and MCP

Every automation surface talks to the running app, so scripts and coding agents use the configured models and profiles without handling API keys.

Terminal
# Transcribe an audio file to the clipboard
$ speak transcribe memo.m4a | pbcopy

# Dictate a commit message
$ speak listen
$ git commit -m "$(speak stop)"

# Register the MCP server for Claude Code
$ claude mcp add justspeaktoit -- speak mcp

Local socket access stays off until enabled in Settings → General → Automation.

  • CLI (speak) transcribe, listen, stop, history and status, with versioned --json output and distinct exit codes so scripts can branch cleanly.
  • MCP server (speak mcp) Exposes dictation, file transcription and History over stdio to Claude Code and other MCP clients, with idempotent start and stop calls.
  • Shortcuts and URL scheme App Intents on Mac and iOS, plus justspeaktoit:// and x-callback-url actions when the iOS app is in front.

Automation guide →

Download

Download for every device

Free and MIT licensed, with no account to create. Mac builds are stable releases; iPhone, iPad, Linux and Windows builds are public betas.

macOS

Stable · v3.2.0

macOS 14 or later. The direct download uses macOS Accessibility to type straight into active fields (there is no Mac App Store build).

Download for Mac

Detected build: Universal compatible · Apple Silicon · Intel (universal)

Homebrew: brew install --cask crmitchelmore/justspeaktoit/justspeaktoit