Push to Talk for Claude Code: What It Takes

The Whisper Key ($87.90) is a handheld unit you hold near your mouth, with a key you press while you speak. Before anything else, the honest part: it does not do speech recognition. It is a microphone and some keys. The words become text because of whatever dictation software you already run. That limitation is also the reason it works with Claude Code at all.

Does Claude Code have a push to talk button?

No. Claude Code reads typed text in a terminal, and it has no idea whether those characters arrived from your keyboard, a paste or a dictation engine. So there is nothing to enable inside Claude Code and nothing that can break there either. What people mean by push to talk for Claude Code is push to talk for their operating system's dictation, aimed at whichever terminal window has focus.

Why hold a key instead of leaving the microphone open?

A Claude Code session is long and spiky. You brief it, it goes away and reads files, it comes back with a question, you answer, then you approve or reject a change. Between those moments there are stretches where you are not talking to the agent at all. You may be on a call, or answering the person next to you, or reading something out loud to yourself.

An always listening microphone hears all of that and tries to turn it into text. A key you hold turns the microphone into something you aim. Hold, speak, release. The room stays out of your prompt, and you never have to check a screen indicator to find out whether you are live, because your thumb already knows.

What does the hardware actually do?

Two things. It puts a microphone near your mouth, and it gives you a key to hold while you talk. The Whisper Key is handheld, so the microphone goes where your hand goes, which is the shape that makes sense in a room with other people in it. The Voice Pad ($79) sits on the desk instead and is supplied with a microphone, three keys and a knob. The Vibe Pad ($59) is a three key pad sold with a separate clip on microphone, so you choose where the microphone ends up. In all three cases the recognition happens in software you picked.

What should the other keys do?

Whatever your review loop needs, because these pads send ordinary keystrokes. A sensible starting map is one key held for dictation, one that sends Enter to submit what you dictated, and the rest set to whatever keystrokes you already use to accept or decline a proposed edit. The pad does not know Claude Code exists. You are moving keystrokes you already press off your main keyboard and onto something you can reach without looking.

Who should skip this?

Two groups. If you work alone behind a door and typing prompts is not your bottleneck, you are buying a solution to a social problem you do not have. And if your dictation software only offers a toggle rather than a hold, push to talk is not available to you no matter what hardware you buy, so check that before you spend anything. We have not confirmed behaviour for every dictation tool, and it is the one thing worth testing first.

If you have got that far and want the practical order of operations, read how to set up push to talk for dictation.