Claude

Use case · Updated July 2026

Claude voice input in 2026: what Claude.ai and Claude Code offer natively, and how to dictate with FluidVox

"Claude voice input" actually points to three different features: Claude.ai's conversational voice mode, Claude mobile's speech-to-text dictation, and Claude Code's /voice for the terminal. Each is genuinely useful, and each has a real limit — conversation-only, mobile-only, or cloud-only. This page covers what each one actually does, then how FluidVox fills the gaps: dictating prompts into Claude on the web or desktop, and dictating straight into Claude Code's terminal, on Mac and Windows.

Download for macOS

Native features

Does Claude already have voice input built in?

Claude.ai voice mode — conversation, not a composer

Voice mode is available on the web and in the Claude mobile apps, on every plan including free, and now supports multilingual input in beta. It's a genuine two-way conversation: you speak, Claude speaks back, with hands-free and push-to-talk options. What it isn't built for is typing text into a document, email, or someone else's app — it's a chat interface, not a dictation tool, and the transcript lives inside your Claude conversation history rather than landing at your cursor elsewhere.

Claude mobile dictation — real speech-to-text, but iOS/Android only

This is the feature closest to true dictation: tap the microphone icon beside the chat input in Claude for iOS or Android, speak, and your words convert to editable text in the prompt box. Anthropic says the audio is deleted after transcription and isn't used for training. It's free, it works well for its job, and it's limited to that one text field inside Claude's own mobile app — there's no equivalent for Claude on the web or desktop, and no way to dictate into other apps with it.

Claude Code's /voice — solid for the CLI, cloud-only, no SSH

Developers searching "claude code voice input" are usually looking for this one. Run /voice in the Claude Code CLI (v2.1.69+) to enable hold-to-record or tap-to-record dictation, with transcription tuned for coding vocabulary — terms like regex, OAuth, and localhost come through cleanly, and your project name and git branch are added as recognition hints automatically. It also works in the Claude Code VS Code extension.

The catches: it requires signing in with a Claude.ai account rather than an API key, Bedrock, Vertex AI, or Foundry; audio streams to Anthropic's servers rather than staying on-device; and it doesn't work in remote sessions — SSH, Claude Code on the web, or VS Code Remote — because the microphone has to be local to where the CLI is running.

Where FluidVox fits for Claude Code

FluidVox isn't a Claude Code plugin — it's a system-wide dictation app that happens to work in your terminal the same way it works in every other window. Hold your FluidVox hotkey with a local Terminal, iTerm2, or VS Code integrated terminal focused, speak your instruction, and it lands at your cursor — whether Claude Code's own /voice is enabled or not, and regardless of whether you're signed in with an API key.

On the Local plan, that dictation runs entirely on-device rather than streaming to a server. FluidVox's Code style tightens punctuation and formatting for exactly this kind of input. It doesn't solve the remote-session limitation either — audio capture is inherently a local-machine problem for any dictation tool, Claude's included.

Side-by-side

Claude's voice features vs FluidVox

Claude.ai voice mode Claude mobile dictation Claude Code /voice FluidVox
What it's for Spoken conversation with Claude Speech-to-text into the mobile prompt box Speech-to-text into the CLI prompt Speech-to-text into any app, including Claude
Platforms Web, iOS, Android iOS, Android only Terminal + VS Code, local machine only Mac, Windows, iPhone
Works in Claude web/desktop text fields N/A — it's the conversation itself No No Yes
Runs on-device / offline No No No — streams to Anthropic's servers Yes — Local plan is fully on-device
Account requirement Any Claude account, all plans Any Claude account Claude.ai account only (no API key/Bedrock/Vertex/Foundry) FluidVox account, independent of Claude
Works over SSH / remote sessions N/A N/A No No — local sessions only (same audio-hardware limit)
Custom vocabulary / dictionary No No Auto-hints from project/branch name only Yes — personal dictionary, auto-learned corrections
Per-app tone/style control No No No Yes — 6 styles, matched automatically per app
Cost Free, all plans Free Free, doesn't count toward usage limits 14-day free trial, then $2.99/mo or $39 one-time

None of this is a knock on Claude's voice features — mobile dictation and Claude Code's /voice are both well built for what they do. The honest gap is that neither one follows you outside its own text field, and neither runs offline. That's the specific space FluidVox occupies.

How it works

Dictating into Claude with FluidVox

1

Open Claude

The web app, the desktop app, or a Claude Code session in your terminal — FluidVox works the same way in all three.

2

Hold your hotkey and speak

Fn on Mac, Ctrl+Shift+Space on Windows (both configurable). Fn+Space or Ctrl+Shift+H toggles hands-free mode for longer prompts.

3

Release and review

Formatted text appears at your cursor, in the style matched to whatever app you're dictating into.

Smart Formatting

Detailed prompts, naturally

Speaking tends to produce more context and detail than typing the same request — useful for prompts, less useful when you actually wanted three words. FluidVox's Natural style keeps conversational phrasing intact for prompt boxes; the Code style tightens things up when you're dictating into a terminal or a code comment instead.

Claude
I have a Node.js Express API that's experiencing memory leaks in production. The heap grows by about 50MB per hour under normal load. Can you help me identify common causes and suggest a debugging approach? The app uses Mongoose for MongoDB, Redis for caching, and handles about 1000 requests per minute.
Natural conversational style

Speed

Explain faster than you type

Speaking is naturally faster than typing, and FluidVox transcribes it in real time. That matters most for the long, context-heavy prompts Claude tends to handle best — architecture questions, debugging walkthroughs, document review requests.

Voice with FluidVox
~150 WPM
Keyboard typing
~50 WPM

Accuracy

Project terms preserved, everywhere you dictate

Claude Code's /voice auto-hints your current project and branch name — but only inside Claude Code. Add the same terms to FluidVox's personal dictionary once and they're recognized correctly in Claude Code, the Claude web app, Slack, email, and everywhere else you dictate.

Custom Dictionary
TypeScript TypeScript
PostgreSQL PostgreSQL
OAuth OAuth

A day dictating into Claude with FluidVox

9:00 AM

Terminal, Claude Code

Dictate a bug description straight into your terminal, technical terms preserved via your dictionary

11:00 AM

Browser, document review

Speak a long prompt asking Claude to review and summarize a document — hands-free mode for the length

2:00 PM

Desktop app, architecture

Talk through your system design in the Claude desktop app for Claude to evaluate trade-offs

4:00 PM

Writing review

Dictate feedback criteria for Claude to apply while reviewing a draft, in Professional style for the write-up after

Claude voice input FAQ

Start dictating into Claude today

Install FluidVox once and it works in Claude, Claude Code's terminal, and every other app on Mac or Windows. 14-day free trial, no credit card.

Download for macOS