Suggestion: Voice input

If you're on a mac in iterm2 you can try /listen

@8enmann how is it working?

> /listen

⏺ Listening intently for your next instructions...

> /stop

⏺ I've stopped listening. Ready for your next instructions or request.

And nothing happened.

Yet made own python script for iTerm2 with speech record while 'fn' is pressed

I'm on Linux 🐧 😣

I'd love to be able to configure this with a push-to-talk hotkey rather than having to type a command each time (to minimise latency).

@8enmann how is it working? `` > /listen ⏺ Listening intently for your next instructions... > /stop ⏺ I've stopped listening. Ready for your next instructions or request. `` And nothing happened. Yet made own python script for iTerm2 with speech record while 'fn' is pressed

It uses mac built-in dictation. There's no /stop command, just hit escape when you're done talking.

Claude Code works with built-in system dictation. Just enable it in accessibility settings, and speak. This is how I do a lot of my Claude Code work!

@bcherny +1, but how do you not annoy your coworkers/wife etc? Separate office?

I can't seem to see this /listen command available, and can't see any other mention of it elsewhere. Is it documented somewhere official to find out exactly how to enable it?

We no longer have this unfortunately. We recommend using the built in
dictation in your OS (that's what /listen did anyway).

On Tue, Jun 24, 2025 at 7:35 PM Jake Shelby @.***>
wrote:

jmshelby left a comment (anthropics/claude-code#154) <https://github.com/anthropics/claude-code/issues/154#issuecomment-3002510180> I can't seem to see this /listen command available, and can't see any other mention of it elsewhere. Is it documented somewhere official to find out exactly how to enable it? — Reply to this email directly, view it on GitHub <https://github.com/anthropics/claude-code/issues/154#issuecomment-3002510180>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AADSWQQMR6TF6IGHKORUYC33FIDHTAVCNFSM6AAAAABX6I4H6CVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZTAMBSGUYTAMJYGA> . You are receiving this because you were assigned.Message ID: @.***>

Hey @gerrywastaken! 👋

I saw this issue was closed with the OS dictation recommendation, but wanted to share that we've built something that goes way beyond basic dictation - true voice conversations with Claude Code!

Voice Mode is an MCP server that provides:

Beyond dictation:

• Two-way voice conversations - Not just input, but hear Claude's responses too
• Smart recording - Automatically stops when you pause (no manual stop needed)
• 3x faster than typing - As you mentioned, but even better when you hear responses!

Quick setup:

# One command setup
claude mcp add voice-mode --env OPENAI_API_KEY=your-key -- uvx voice-mode

# Start conversing
claude converse

Then tell Claude: "Let's have a voice conversation" and you're off!

Why it's better than OS dictation:

1. Natural conversations - No push-to-talk, just speak naturally
2. Context-aware - Claude can ask clarifying questions verbally
3. Code-optimized - Handles technical terms and code syntax well
4. Cross-platform - Works on Linux, macOS, and Windows (WSL)

The push-to-talk idea from @djrodgerspryor is interesting - we use automatic silence detection instead, which feels more natural for conversations.

Would love to hear if this solves the productivity boost you were looking for!

Interesting!

The push-to-talk suggestion is 90% about not picking up irrelevant conversation (I'm not always working alone, often on phone calls etc.). I personally don't want an always-listening approach.

I created mbailey/push2type because I couldn't find what I wanted online.

I use mbailey/voicemode for Claude Code and push2type for other apps.

Oh, and justsayit (addon for firefox) for web.

push2type

Simple push-to-talk voice transcription for Linux. Hold CapsLock to record, release to transcribe and paste text anywhere.

Features

• 🎤 CapsLock as push-to-talk - Hold to record, release to transcribe
• 🔤 Accurate transcription - Uses OpenAI Whisper API
• 📝 Auto-paste - Text appears where you're typing
• 💾 Save recordings - Audio and text files with timestamps
• 🖥️ Works everywhere - X11 and Wayland support
• ⚡ Simple & reliable - Pure bash, no complex dependencies

Quick Start

1. Run setup:

``bash
# Clone and run
git clone https://github.com/mbailey/push2type.git
cd push2type
./setup.sh
``

Or run directly:
``bash
curl -sSL https://raw.githubusercontent.com/mbailey/push2type/master/setup.sh | bash
``

This will install all required dependencies and configure your system.

1. API key configuration:

The setup script will prompt for your OpenAI API key and save it to ~/.config/push2type/environment.

If you need to update it later:
``bash
echo "OPENAI_API_KEY=your-openai-api-key" > ~/.config/push2type/environment
chmod 600 ~/.config/push2type/environment
``

1. Start the daemon:

```bash
# As systemd service (recommended)
systemctl --user start push2type

# Or run directly
push2type
```

1. Use it:

• Hold CapsLock to record
• Release CapsLock to transcribe and paste

This issue has been automatically locked since it was closed and has not had any activity for 7 days. If you're experiencing a similar issue, please file a new issue and reference this one if it's relevant.