Dictate into
any Mac app

Hold a key, talk, and let go. InkIt puts your words at the cursor in any Mac app, from your editor to email or the terminal. It streams your voice to Cartesia Ink 2 on your own API key.
#launchMessages
Inserted
Sam 10:42

Did the launch date change?

Hey Sam, quick update. The launch moved to Thursday, so let's review the copy tomorrow.
InkItPolished

Example run. A scripted example of the app at work, not a recording.

Runs on
macOS
Model
Ink 2
License
Apache-2.0
Needs
A Cartesia API key
Stack
macOSSwift
Cartesia products
Ink 2

What it does

InkIt is free and open source under Apache-2.0, with no subscription. You pay only for the speech to text you use, on your own Cartesia key. It pastes through macOS accessibility access and puts your clipboard back afterwards, so there is nothing to set up per app. Dictation is English only for now. Every dictation is saved in a searchable history.

Run it yourself

  1. Download and open InkIt

    Move it to Applications and open it. A short setup asks for microphone access and for accessibility access, which macOS requires before an app can paste into other apps.

  2. Paste a Cartesia API key

    Create a free key at play.cartesia.ai. It covers about 15,000 words of dictation a month. Paste it in when setup asks.

  3. Pick a shortcut and talk

    By default you hold fn to talk and let go to paste. A tap on the right Command key starts hands-free dictation, and a second tap stops it. Change either key in the Dictation settings.

Build it with your coding agent

One prompt that tells your coding agent where to start and what to read first.

Prompt

Claude CodeCursorCodex

Fork https://github.com/cartesia-ai/InkIt and add a feature to it. Read the README and AGENTS.md first. InkIt is a Swift macOS app built with XcodeGen that streams microphone audio to Cartesia Ink 2 and types the transcript at the cursor.

How it works

MicHold your shortcut
Ink 2Streams text as you speak
PolishOptional cleanup by an LLM on your key
Your appText appears at the cursor

Why it works this way

It pastes for you, then puts your clipboard back

InkIt puts the text on the clipboard and presses Command-V for you, so it works in any app that accepts a paste. It saves what you had copied first and restores it 0.4 seconds later, unless something else has replaced it by then. The pasted text is tagged so clipboard managers don't save it.

Text lands once, after you let go

Ink 2 sends partial text while you talk and revises it as it hears more. When you let go, InkIt asks Ink 2 to finish, waits up to 3 seconds for the final text and pastes it in one go. If no text field has focus by then, the text goes to History instead.

Polish cleans up but does not rewrite

The prompt in TranscriptRewriter.swift tells the model to drop filler like 'um' and repeated words, and to keep only the corrected version when you say 'scratch that' or 'I mean'. It copies numbers exactly and never answers questions you dictate. If the model fails or returns far more text than you said, InkIt pastes the raw transcript.

Make it yours

Your vocabulary

Add product names, people and spellings under Dictionary once. The list goes along with every dictation, so Ink 2 listens for those words and spells them your way.

Polish with your own model

Turn on Polish transcripts in settings. Add a key from Groq, Gemini, OpenAI or Anthropic. Before pasting, InkIt sends the transcript to that model to fix punctuation and misheard names and drop filler words. It stays off until you add a key.

Fork it

Build it with Xcode using the commands in the Code tab. Then change the shortcuts, the history or the cleanup prompt. The prompt lives in TranscriptRewriter.swift.

Questions

How much does InkIt cost?

The app is free. Dictation runs on your Cartesia API key, and a free key covers about 15,000 words a month. Polish, if you turn it on, uses your own LLM key.

Does InkIt work offline?

No. InkIt sends audio to Cartesia Ink 2 to transcribe it, so it needs an internet connection.

How is InkIt different from built-in macOS dictation?

It transcribes with Cartesia Ink 2. It also takes your own word list, keeps a searchable history and can remove filler words with an LLM you pick.

Does InkIt keep my recordings?

Only on your Mac. It keeps the audio of your last 50 dictations from the past week so you can attach one to an issue report, and clearing your history deletes them. Audio leaves your Mac only to be transcribed by Cartesia, or in a report you choose to send.

Which Macs does InkIt run on?

InkIt needs macOS 14 Sonoma or later.

Get started today

Talk to an expert.

Connect with a member of our team and learn how Cartesia can help you build world-class voice experiences.

Contact Sales

Start building.

Access our models via API and bring a voice agent into production in minutes.

Try Cartesia