Talk to your terminal. Claude Code listens. Voice dictation for CLI ▎ Whisper transcription, rule-based prompt optimizer, auto-paste. macOS.

[nodox-studio]

Display Name

claude-voice

Category

Tooling

Sub-Category

General

Primary Link

https://github.com/nodox-studio/claude-voice

Author Name

Guille Varela

Author Link

https://github.com/nodox-studio

License

MIT

Other License

No response

Description

Voice dictation tool for Claude Code that records microphone input, transcribes it locally using Whisper (offline, no data sent to cloud), and pastes the optimized text directly into the Claude Code terminal input.

A rule-based optimizer strips filler words, removes polite preambles, and normalizes verb forms into direct imperatives. Supports Ghostty split-panel mode via a /voice slash command.

Validate Claims

Install takes under 2 minutes:

• Clone the repo, run ./install.sh, then type /voice inside Claude Code.
• Speak a sentence with filler words ("uh can you like fix the bug in the login form") and press 1
• — the optimized text ("Fix the bug in the login form.") appears in the Claude Code input.
• No API key, no account, no configuration required beyond granting microphone and Accessibility permissions on first run.

Specific Task(s)

1. Type /voice in Claude Code.
2. Speak: "uh can you like refactor the authentication component, and also add tests, and also update the docs"
3. The tool detects the multi-part request and automatically structures it as a numbered list:

Output:

• 1. Refactor the authentication component.
• 2. Add tests.
• 3. Update the docs.

4. Press 1 — the full list is pasted directly into the Claude Code input with a
   single keypress.

Specific Prompt(s)

After installing, type /voice in Claude Code and speak exactly this sentence: "uh can you like refactor the authentication component, and also add tests, and also update the docs" — then press 1 to send. No prompt to Claude Code is needed — the tool handles everything.

PressCtrl+Cto exit the tool or Cmd+W to close the split panel.

Additional Comments

Network requests: Whisper runs fully offline.
The only optional network request is a fallback to Google Speech API if Whisper fails with a technical error — the tool displays an explicit warning on screen when this happens. It can be disabled by removing the fallback block in voice_prompt.py.

Uninstall: delete ~/.local/bin/voice, ~/.claude/commands/voice.md, ~/.claude/scripts/voice_prompt.py, and the venv at the repo's .venv/ directory.

Recommendation Checklist

• I have checked that this resource hasn't already been submitted
• It has been over one week since the first public commit to the repo I am recommending
• All provided links are working and publicly accessible
• I do NOT have any other open issues in this repository
• I am primarily composed of human-y stuff and not electrical circuits

[hesreallyhim]

Thanks for the recommendation! This repository is less than a week old. We ask that projects have some time in the wild before being recommended — you're welcome to re-submit after May 26, 2026.

A cooldown of 7 days has been applied.

[nodox-studio]

Understood, will resubmit after May 26. Thanks for the quick response.