Voice dictation lets you speak your prompt instead of typing. Hold a key, talk, release, and the transcription fills the prompt buffer.

What it does

Claude Code records audio while you hold the dictation key, transcribes it locally or via Anthropic's speech service depending on config, and inserts the text at your cursor. You can pick a language, pick a model, and edit the result before submitting. It works on macOS, Linux, and Windows.

When to use it

Long, ranty prompts that are faster to say than type.
Accessibility workflows where typing is tiring or not possible.
Brainstorming out loud while Claude captures the intent.
Mobile-adjacent setups where you're away from a full keyboard.

Gotchas

Microphone permission has to be granted to your terminal emulator. Some terminals need a restart after the grant.
Background noise matters. Transcription quality drops fast in a noisy room.
The transcript lands as plain text - double-check code snippets or file paths before submitting.

Official docs: https://code.claude.com/docs/en/voice-dictation.md