Voice dictation for Claude Code lets you speak prompts instead of typing them, and Claude Code now ships a built-in option: run /voice, hold Space, and your words appear in the terminal. It’s tuned for coding terms but needs a Claude.ai account, streams audio to Anthropic’s servers, and won’t work over SSH or a bare API key.
The gap between talking and typing is bigger than most developers assume. Typing a detailed prompt tops out around 40 words per minute for most people once you count thinking pauses and edits. Speaking the same prompt runs closer to 150 words per minute in natural conversation — nearly four times faster. For a 200-word prompt describing a refactor, that’s roughly five minutes of typing against under ninety seconds of talking. Learning to dictate prompts to Claude Code well is mostly a matter of structuring what you say, not just talking faster.
Pairing an AI coding agent with Replit too? Check out voice dictation for Replit for Agent-specific voice prompting tips and a spoken prompt cheat sheet.
Typing vs Speaking: Why Long Claude Code Prompts Slow You Down
Claude Code rewards detail. A vague prompt like “fix the auth bug” gets a worse result than one that names the file, the expected behaviour, and the edge case that’s failing. Typing all of that is where most developers give up and write the short, lazy version instead — then spend three follow-up messages correcting Claude’s guesses.
Talking removes that friction. You can describe a stack trace, list acceptance criteria, and mention file paths in one breath, the way you’d brief a colleague. That’s the real case for claude code voice input: getting full context into the prompt the first time instead of trickling it out over five exchanges.
Claude Code’s Built-In Voice Dictation (/voice)
Anthropic added native dictation to Claude Code, and it’s a genuinely useful first step if you live entirely inside the CLI. Run /voice to turn it on, then hold Space to record push-to-talk style (the default), or switch to tap mode with /voice tap to start and stop with single key presses. Speech streams into the prompt live, and you can mix typed and spoken text in the same message.
Transcription is genuinely tuned for code: terms like regex, OAuth, JSON, and localhost come through correctly, and Claude Code feeds in your current project name and git branch as recognition hints automatically. Dictation also doesn’t burn Claude messages or tokens. Anthropic’s own documentation is the definitive reference for the full command list.
The limitations matter as much as the features. Voice dictation streams your recorded audio to Anthropic’s servers — nothing is processed locally. It requires a Claude.ai account: it does not work if Claude Code is configured with a direct API key, Amazon Bedrock, Google Cloud’s Agent Platform, or Microsoft Foundry. And because it needs a live local microphone, it doesn’t work in remote setups — Claude Code on the web, SSH sessions, or a VS Code Dev Container or Codespace.
How to Set Up /voice in Claude Code
Most people searching for this want to know how to dictate prompts Claude Code will execute correctly on the first try. The setup below gets you there in under a minute:
- Update to Claude Code v2.1.69 or later, then sign in with /login using a Claude.ai account (not an API key).
- Type /voice in the CLI and allow the microphone prompt in System Settings (macOS) or Settings → Privacy → Microphone (Windows).
- Choose your mode: /voice hold for push-to-talk, or /voice tap for tap-to-start/tap-to-stop with no warmup delay.
- On WSL, install WSLg (bundled with WSL2 from the Microsoft Store) or Claude Code won’t find a working audio device.
- Optionally set “autoSubmit”: true in the voice settings block so a finished transcript sends itself automatically.
Dictation persists across sessions once configured. To change the listening language, set the language value in /config — English is the default.
Where /voice Runs Out of Road
The built-in option is good at what it covers, but that’s narrow: the Claude Code CLI on a local machine with a Claude.ai login. The moment your day moves to a remote terminal, a browser tab, Slack, Gmail, or a GitHub PR, /voice goes silent and you’re back to typing.
That’s the real gap for most developers: you don’t just want to talk to Claude Code, you want one dictation habit across the terminal, Claude.ai in the browser, your IDE, GitHub, and Slack. There’s also the raw-transcript problem — /voice transcribes exactly what you said, filler words and false starts included, rather than cleaning rambling into a tight prompt.
Genie 007: Voice Dictation for Claude Code That Works Everywhere Else
This is where Genie 007’s think-to-text approach earns its keep. Instead of transcribing every “um” and false start literally, Genie Mode listens for what you meant and delivers a clean, structured prompt — so a rambling bug explanation becomes a precise instruction with the file, symptom, and fix criteria laid out properly. Voice Typing mode handles faster, more literal dictation with automatic punctuation when that’s what you want instead.
Genie 007 ships on four platforms — Windows, Mac, mobile, and a browser extension — not as a single-app add-on. The native Windows and Mac apps work system-wide, including your terminal, so the same shortcut that drafts a Claude Code prompt also works in your IDE, Slack, Gmail, and GitHub PR descriptions, without switching tools between contexts. Unlike /voice, it doesn’t require a Claude.ai account at all.
The other honest differentiator is language: 140+ languages with live translation, so a non-native English speaker can talk through a problem in their first language and get a polished English prompt out the other end — a workflow /voice, Apple Dictation, and Windows Voice Access don’t match.
Setting Up Genie 007 for Claude Code Prompts
- Install the Windows app or Mac app for system-wide dictation in your terminal and IDE, or add the browser extension for Claude.ai and other web tools.
- Start the free trial (no credit card) and grant microphone access when prompted.
- Trigger the hotkey in your terminal and describe the change you want in plain language — file, bug, and what “done” looks like.
- Let Genie Mode clean the rambling into a structured prompt, review it, and send it into Claude Code.
- Reuse the same hotkey in your browser, Slack, or GitHub without reconfiguring anything.
None of this requires switching accounts or apps between contexts — the same hotkey and the same cleaned-up prompt style follow you everywhere.
Coming from Cursor rather than the CLI specifically? The setup is nearly identical — see the Cursor voice dictation guide, or the broader vibe coding voice dictation workflow for describing whole features rather than single fixes.
Voice Dictation Tools for Claude Code Compared (2026)
Prices, platform coverage, and where each tool actually works:
| Tool | Price/month | Free plan? | Platforms | Works in | Best for |
|---|---|---|---|---|---|
| Genie 007 | £5–£10 (~$7–$14) | Yes, free trial, no card | Windows, Mac, mobile, browser extension | Terminal, IDE, Gmail, Slack, GitHub, browser — system-wide | The one dictation habit that covers Claude Code and everything else around it, in 140+ languages |
| Claude Code built-in /voice | Free (included) | Yes, no separate plan | macOS, Windows, Linux (WSLg), VS Code extension | Claude Code CLI and VS Code panel only, local machine | Developers who live entirely inside the Claude Code CLI with a Claude.ai login |
| Apple Dictation | Free | Yes, built-in, unlimited | Mac, iOS, iPadOS only | Any native text field, including Terminal — Apple devices only | Mac/iOS users who want zero-cost, on-device dictation and don’t need coding vocabulary |
| Windows Voice Access | Free | Yes, built-in, unlimited | Windows 11 only | Any text field, including Terminal — Windows devices only | Windows users who want free, basic voice control with no AI cleanup and an English-focused command set |
| Dragon Professional | ~$16.63 (or ~$699 one-time) | No free plan | Windows only (legacy) | Any Windows desktop text field, including Terminal | Enterprise users needing specialized vocabularies at a high price, without modern AI cleanup |
Every other tool wins somewhere narrow. If you never leave the Claude Code CLI and don’t mind your audio going to Anthropic’s servers, /voice costs nothing extra. If you want offline, zero-cost dictation and live fully in the Apple ecosystem, Apple Dictation is hard to beat on price. Windows Voice Access is a free way to get basic voice control built into Windows 11, though it’s a command set rather than AI-cleaned dictation. Dragon Professional is the legacy choice for specialized vocabularies — medical, legal, and other niche terminology — but it’s Windows-focused, priced well above the rest, and doesn’t do modern AI cleanup. Genie 007 is the one built for breadth: one subscription, four platforms, 140+ languages, and a prompt that’s cleaned up rather than transcribed literally — which is why it’s the overall pick for developers who want a single dictation habit everywhere they work, not just in Claude Code.
One habit helps regardless of tool: name the file path and function early in the sentence, since both /voice and Genie 007 use that context to bias transcription toward the right terms. For adjacent setups, see the VS Code voice dictation and GitHub voice dictation guides, or dictating to Claude.ai for the web-chat side of planning work.
Frequently Asked Questions
Does Claude Code have voice dictation built in?
Yes. Claude Code added native voice dictation, enabled with the /voice command. It uses push-to-talk (hold Space) by default or a tap-to-record mode, transcribes speech live into the prompt, and is tuned for coding terms like regex, OAuth, and localhost.
How do I enable voice mode in Claude Code?
Update to Claude Code v2.1.69 or later, sign in with a Claude.ai account using /login, then run /voice in the CLI. Grant microphone access when prompted, and choose hold mode or tap mode with /voice hold or /voice tap.
Does Claude Code voice dictation work over SSH or on a remote server?
No. Claude Code’s built-in /voice requires a local microphone and does not work in remote environments such as Claude Code on the web, SSH sessions, or the VS Code extension running in a Dev Container or Codespace.
Is Claude Code’s voice input private?
Not fully private by default: /voice streams your recorded audio to Anthropic’s servers for transcription rather than processing it on-device, and it requires a Claude.ai account rather than an API key, Bedrock, or Vertex AI setup.
What is the fastest way to dictate prompts to Claude Code?
For CLI-only work, Claude Code’s built-in /voice is fastest to set up since nothing extra installs. For dictating across the terminal, browser, and every other app you use, a system-wide tool like Genie 007 avoids switching dictation methods between contexts.
If you use Windsurf alongside Claude Code, the same think-to-text workflow applies — see the full guide to voice dictation for Windsurf.
Try Genie 007 Free
If /voice has you sold on talking to Claude Code but you want the same habit in your terminal, browser, and every other app on your desk, Genie 007 is built for exactly that.
Download Genie 007 free — available for Windows, Mac, mobile and as a browser extension. No credit card required.
Written by Bill Kiani, founder of Genie 007.
If you prompt both Claude Code and an AI app builder by voice, our voice dictation for Lovable guide covers the equivalent workflow for Lovable.
If you use v0 alongside Claude Code in your vibe-coding workflow, see our guide on voice dictation for v0 — the same Genie 007 setup covers both tools from a single shortcut.
If you split your day between Copilot and a coding assistant, the same hotkey habit covers copilot dictation across Word, Outlook and Edge too — read the full guide.
