The dictation app
built for AI.

Speak your prompts, chat from the notch, and circle anything on screen to ask about it or translate it. Your words land right where you’re typing.

See Cadence in action
  • OpenCode
  • Antigravity
  • Claude
  • OpenAI / Codex
  • Grok
  • Cursor

Meet Cadence.

Try it · AI coding tools

See it in Claude Code.

Hold to describe a bug out loud and watch Cadence drop a clean prompt into Claude Code. For a bigger task, talk it through with Refine, review the prompt Cadence drafts, then insert it. The same two moves work in your other AI coding tools.

✻ Welcome to Claude Code
/help for help, /status for your current setup
cwd: ~/code/acme-dashboard
Tips: ask Claude to fix a bug, write a test, or explain the code.
The settings page crashes when the API key is empty. Can you fix that and add a test for it?
RewriteTerminal
New text at your cursor
What would you like to write?
Presets CopyRewrite
Inserted

“Um, so the settings page crashes when the API key is empty. Can you, uh, fix that and add a test for it?”

Interactive example with a sample voice. Nothing is recorded. Claude Code and the terminal are recreated; the pill and Refine panel follow the real Cadence app.

Speak into your favorite AI tools
ChatGPTClaudeGoogle AntigravityCursorDevin
You said Filler and false starts removed Capitals and numbers fixed Inserted at your cursor

Um,canCanwemovethemeetingtoThursday,actuallyFriday,Fridayatten?10?

AI Enhancement is optional. Turn it on and every dictation is tidied before it lands, while keeping what you meant.

Removes filler and repetition
Improves punctuation and capitalization
Applies spoken corrections

Better prompts.
In your own words.

Highlight a rough prompt and tell Cadence how to improve it. Make the task clearer, organize your instructions, or trim the repetition. Review the rewrite before you press Accept.

Try turning this rough idea into an AI prompt. This is an interactive example.

What would you like to build?
AI coding chat · example
make a dashboard with search and filters and make it work on mobile
Tell Cadence what to change…

Chat

Ask anything
from the notch.

Chat opens from the notch at the top of your screen, over whatever you’re doing. Attach PDFs, Word documents and spreadsheets, ask follow-ups, and open the full Chat window when you want more room.

Files are read on your computer, and only their text is sent.

Cadence Chat open in the notch: two attached files, a summary with a small table, and a reply box

Try it · Annotate and Translate

Circle it.
Ask about it.

Press the Annotate shortcut and circle anything on screen: a quiz question, a paragraph, a note in another language. Pick Answer it, Explain, Summarize or Translate, or ask your own question. The answer opens in Chat, right from the notch.

Go on, circle something on the page below.

BIO 101 · Week 4

Cellular respiration

Reading · 4.2

Cellular respiration is how cells release the energy stored in food. It starts in the cytoplasm, where glycolysis splits each glucose molecule into two molecules of pyruvate. The pyruvate then moves into the mitochondria, where the Krebs cycle and the electron transport chain use oxygen to make most of the cell’s ATP. Carbon dioxide and water are left over as waste.

Quiz · Question 3 of 10

Which organelle produces most of the cell’s energy?

  • A) Nucleus
  • B) Mitochondria
  • C) Ribosome
  • D) Golgi apparatus

Time left 12:40

Prof. MoralesAnnouncement · 9:12 AM

Recuerden: el examen del viernes cubre los capítulos 4 y 5. Traigan calculadora y lleguen diez minutos antes. ¡Mucha suerte!

Circle anything to ask Cadence Esc to cancel

Reading the circled area…

Chat
Circled on screenText read on your computer

Cadence can make mistakes.

Translated to

Escto close

Interactive example. The answers and translations are written in advance, and nothing you type is sent anywhere. The browser and course page are recreated; the circle, ask bar, Chat and Translate card follow the real Cadence app. Annotate is ⌃⌥C on Mac and Ctrl+Alt+C on Windows; Translate is ⌃⌥T and Ctrl+Alt+T, with 108 languages.

For everything you write in a day.

CodexClaude Code

AI prompts

Dictate

um add a search bar to the dashboard, let me filter by project and keep the current design

Add a search bar to the dashboard with project filters. Keep the current design.

Gmail

Email

Dictate

um following up on the invoice from last week, can you let me know when it’s been processed

Following up on last week’s invoice. Could you let me know once it’s been processed?

Messages

Messages

Dictate

running like ten minutes late, order me the usual

Running about ten minutes late. Order me the usual?

Apple Notes

Notes

Dictate

idea for the talk, open with the story about the first prototype, then the three lessons

Talk idea: open with the first-prototype story, then the three lessons.

Google Docs

Documents

Enhance · Shorten

The purpose of this document is to outline the various steps that we will need to take in order to launch the product.

This document outlines the steps needed to launch the product.

Tuned to your AI workflow.

Custom shortcuts

Separate keys for dictation and Enhance. Hold to talk, or toggle on and off.

⌥Space

Rewrite presets

Save your favorite ways to refine a prompt, then apply them in one step.

AI promptClarify the taskList requirements

Styles

Save a style for each audience, with example messages, and pick one from the notch. Your dictation comes out the way you’d write it to them.

TeamFriendsClients

Vocabulary hints

Add project names, frameworks, and technical terms as vocabulary hints for your dictation.

One Cadence account

Sign in for managed dictation and rewriting. Check your usage and plan in your account.

Your voice stays yours.

You decide when the microphone is on.

Audio and text are processed to fulfill your request.

Cadence does not retain recordings by default.

Sign in to use Cadence’s managed speech and AI services within your plan’s allowances. Read our privacy policy for processing and retention details.

Bring your voice to AI.

macOS

Apple Silicon · macOS 14 or later

Download for macOS
See release detailsRelease notes

Windows

Windows 10 (build 19041) or later · x64

Download for Windows
See release detailsRelease notes

See download details for the current release and service availability.

The Windows installer is unsigned. View Windows setup and verification notes.

Mac download details · Plans and allowances

Questions.

What is the difference between dictation and AI Enhancement?

Dictation converts your speech to text and inserts it at your cursor. AI Enhancement is an optional step that cleans up that transcript, or rewrites selected text based on your instructions. You can use dictation on its own.

Can I use Cadence for vibe coding and AI agents?

Yes. Dictate a feature brief, debugging request, or task instructions into a supported app’s prompt field. Use Enhance to refine the text, or choose Create AI prompt to turn a spoken idea into a prompt written for your agent. Review it, then send it to your AI tool. Cadence prepares the prompt; the tool you send it to handles the task.

Do I need to set up a provider?

No. Create a Cadence account and sign in inside the app. Managed speech recognition and AI rewriting are included within your plan’s allowances; you do not need provider API keys.

How much does Cadence cost?

Start with a free account. See Plans for the current allowances and Pro availability. You can manage your plan and billing from your account. View plans.

Which platforms are supported?

Cadence is available for Apple Silicon Macs running macOS 14 or later, and Windows 10 (build 19041) or later on x64 PCs. See the download details for the current releases and verification notes.

Can I change the shortcuts?

Yes. The dictation and Enhance shortcuts are both configurable. Choose hold to talk or toggle mode in Settings.

Can I review a rewrite before it is applied?

Yes. Enhance shows a preview of the changed text. Nothing is replaced until you press Accept.

Does Cadence keep my recordings?

Cadence does not retain recordings by default. Audio and text are sent to the services used to process your request. Read the Privacy page for details about processing and retention.

Your next build starts with your voice.