Skip to content

Claude Code Voice Mode: What We Know So Far

Claude Code just shipped /voice — push-to-talk in the terminal. Here's what we know about it.

Claude Code|Share|
Claude Code Voice Mode: What We Know So Far

Claude Code has a /voice command now. Hold space, talk, release. Your words stream into the terminal as text, right where your cursor is. Push-to-talk for code.

It's rolling out gradually — somewhere around 5% of users have access today, with broader availability coming over the next few weeks. I don't have it yet.

What we know

The details so far:

  • Push-to-talk via spacebar. Hold space, say what you need, release. The transcript appears at your cursor position in real time.
  • Hybrid input. You can type half a prompt, voice the rest, then keep typing. Voice and keyboard are interchangeable mid-prompt.
  • No extra cost. Voice mode doesn't add to your bill. Transcription tokens don't count against rate limits either.
  • Available on Pro, Max, Team, and Enterprise plans. Toggle it on with /voice once you have access.

That's most of what's been shared publicly. No word yet on language support, accuracy tradeoffs, or how well it handles technical jargon (function names, file paths, that kind of thing).

Why this is interesting

The obvious use case is dictating long explanations — describing a bug, walking through requirements, giving architectural context. The stuff where typing feels slow and you end up writing less than you mean because it's tedious.

But the hybrid input part is what I'm most curious about. Being able to voice "refactor the auth module in," type src/lib/auth.ts, then voice "to use the new session provider pattern we discussed" — all without switching modes. Voice the intent, type the precise bits like file paths, voice the rest. That could genuinely change how prompts feel.

It also lowers the bar for voice-to-code workflows. Unless you've been using something like Wispr Flow (affiliate link), voice input in a terminal isn't something most developers have had before. Having it built in removes the setup friction entirely.

I'll update this post

I'm in the ~95% who don't have access yet. Once /voice shows up in my terminal, I'll add hands-on impressions — how accurate the transcription is, whether it handles code terminology well, and whether it actually sticks as part of the workflow or ends up being a novelty.

For now, if you're one of the early users, I'd genuinely like to hear how it's working for you.

Happy clauding.

Stay in the loop

Get new posts delivered to your inbox. No spam, unsubscribe anytime.