To dictate to Claude Code, run /voice in the CLI, hold Space, speak, and release; the transcript lands in the prompt and waits for Enter. In Cursor, click the microphone icon in the chat input. Both are the tools' own voice input. For any other terminal or editor chat box, Apple's built-in Dictation or a system-wide push to talk dictation app can type the prompt for you. Whichever you use, read file names, function names and flags before you send.
Publisher note: SuprFlow publishes this guide and is one of the products discussed. Competitor details come from each company's own pages, checked on 1 October 2026. Prices and features change; confirm before you buy.
How does voice dictation work in Claude Code?
Anthropic's Voice dictation page for Claude Code documents it in detail.
- In the Claude Code CLI, run
/voice. Claude Code runs a microphone check; on macOS this triggers the system microphone prompt for your terminal the first time. - Hold Space. The footer shows
keep holding…and thenlistening…. - Speak. Your words appear dimmed in the prompt until finalised.
- Release Space. The transcript is inserted at the cursor, and Claude Code waits for you to press Enter.
/voice hold is this push-to-talk mode and is the default. /voice tap switches to tap mode: tap Space on an empty prompt to start, tap again to stop. /voice off disables it.
What Anthropic's page says it needs:
| Point | What the docs say |
|---|---|
| Account | "A Claude.ai account". It is not available with an Anthropic API key, Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry |
| Where audio goes | "Voice dictation streams your recorded audio to Anthropic's servers for transcription. Audio is not processed locally." |
| Remote sessions | "Voice dictation does not work in cloud sessions or SSH sessions." |
| Languages | Twenty are listed, including English, Spanish, French, German, Hindi and Japanese. Set one in /config; the default is English |
| Key | voice:pushToTalk, default Space, rebindable in ~/.claude/keybindings.json |
| VS Code | The Claude Code VS Code extension supports dictation with the same account requirement |
Two details matter for coding. Anthropic says transcription "is tuned for coding vocabulary", that terms like regex, OAuth, JSON and localhost are recognised, and that "your current project name and git branch name are added as recognition hints automatically."
If you work in Claude Code with a Claude.ai login on your own Mac, this is the best route. It is built in and tuned for the job; Anthropic says transcription "does not consume Claude messages or tokens". It is the wrong route if your policy forbids sending audio to a cloud service, or if you log in with an API key.
Watch tap mode. Anthropic says tap mode "submits the prompt automatically when the transcript is at least three words long." You do not get a review step. Use hold mode, where the default is to wait for Enter, unless you have set autoSubmit.
Does Cursor have voice input?
Yes. Cursor's Prompting agents page has a Voice input section: "Click the microphone icon in the chat input to dictate your prompt instead of typing. Speak naturally, include technical details like file and function names, and review the transcription before sending."
That is all the page says. Cursor's docs do not state where the audio is transcribed, which plans include voice input, or a keyboard shortcut for it as of 1 October 2026.
What about VS Code and other tools?
Visual Studio Code documents built-in dictation, marked experimental. Microsoft says that on desktop it "uses an on-device speech recognition model by default", available on "macOS on Apple silicon". Press ⌘I in the chat input, or ⌥⌘V in an editor; holding the shortcut gives push-to-talk. There is also a Voice: Start Dictation in Terminal command, which Microsoft says "removes ordinary punctuation, converts spoken symbol names, and adjusts initial capitalization" for command-line input.
For a tool with no voice input of its own, use one of the next two routes.
How do you dictate with Apple Dictation?
Apple's Dictation guide for macOS 27 says "you can speak to enter text anywhere you can type it." Turn it on in System Settings → Keyboard → Dictation, click in the prompt, press the Microphone key or your shortcut, and speak. Test it in your own terminal app first.
See How to use voice to text on a Mac, the shortcut guide, the commands list and Mac Dictation not working.
When does a system-wide dictation app fit?
A system-wide app gives you one key for every tool: the terminal, an editor chat box and a browser-based agent. It suits voice to text for coding across several tools, vibe coding included, and anyone who wants speech recognised on the Mac.
SuprFlow is one example. It is a menu-bar app for Apple Silicon Macs on macOS 14 or later. Hold the Fn key, speak, release, and it pastes the text into whichever text field has focus. Speech recognition runs on the Mac after a one-time model download; see https://suprflow.app/requirements. Default formatting is local cleanup of punctuation, capitalization and numbers. It costs $125 once (as of 1 October 2026), with no free tier or trial.
Because it pastes at the cursor, it should work in an ordinary text field; test it in your terminal and your editor's chat box with a harmless sentence. It is a general dictation app, not a coding-tuned one, and we have not run a controlled benchmark against any tool here.
Which route should you use?
| The tool's own voice input | Apple Dictation | System-wide app (SuprFlow as the example) | |
|---|---|---|---|
| Where it works | That tool only | Wherever you can type | Whichever field has focus |
| Start | Claude Code: hold Space. Cursor: click the microphone | Microphone key or shortcut | Hold Fn |
| Speech processed | Claude Code: Anthropic's servers. Cursor: not documented. VS Code desktop: on device | Check Keyboard settings | On the Mac |
| Coding vocabulary | Claude Code documents tuning and project hints | Not documented | Personal dictionary you fill in |
| Cost | Claude Code: no message or token use. Cursor: not documented | Free | $125 once |
| Best for | Staying inside one tool | Occasional prompts | One key across many tools |
Is dictating in a terminal different from an editor chat box?
Yes, in one important way. An agent's prompt box takes plain language, and a wrong word usually costs you a clarifying turn. A shell prompt takes commands, and a wrong word followed by Return runs. Dictate prompts, not shell commands. Make sure the cursor is in the agent's prompt, not a bare shell, before you speak.
Why do file, function and flag names need a review pass?
Speech recognition produces spoken words, and code identifiers are not spoken words. "Get user by ID" is not getUserById, and "dash dash no verify" is not --no-verify.
- Describe, then let the agent find it. "The token validation helper in the auth middleware" survives dictation better than an exact symbol. Cursor's docs say Agent "finds relevant files through its own search."
- Type the exact parts. Dictate the sentence, then type the path, flag or symbol by hand. Claude Code inserts speech at the cursor so you can mix both.
- Use a personal dictionary. Add product names, library names and internal acronyms that keep coming out wrong. See teaching SuprFlow names and acronyms.
What should you check before you press Enter?
- File paths, function names, flags and version numbers.
- Negations: "do not delete", "without changing the schema".
- Branch and environment names, especially anything that says production.
- Anything destructive. If the prompt deletes, force-pushes or migrates, type it.
- Secrets. Never read a key or password aloud.
Questions people ask
How do I turn on push to talk in Claude Code?
Run /voice. Hold mode is the default: hold Space to record and release to insert the text.
Is Claude Code voice dictation processed on my Mac?
No. Anthropic's docs say audio is streamed to its servers and "is not processed locally."
Why does holding Space just type spaces?
Anthropic says voice dictation is likely off, so run /voice hold. If one or two spaces appear and then nothing, key-repeat is not reaching the terminal; try /voice tap.
Can I dictate to Claude Code over SSH?
Not with /voice. The docs say it needs a local microphone and does not work in SSH or cloud sessions.
Related guides
- Dictate to ChatGPT or Claude on a Mac
- How to use voice to text on a Mac
- Mac dictation shortcut
- Teach SuprFlow names and acronyms
- Offline voice to text on a Mac
- Best dictation app for Mac