Voice Mode Comes to Claude Code

Voice mode is beginning to roll out in Claude Code.

It’s currently live for about 5% of users today and will be expanding over the coming weeks.

It’s simple and elegant to use – you hold down the space bar, speak and then release when you’re done. The transcript streams directly at your cursor, allowing you to easily switch between typing and speaking. Now, this isn’t completely new, products like Whispr Flow and Monologue have been doing this for some time (and Claude.ai already has this), but Anthropic have now embedded this functionality directly into Claude Code. It is free within Pro, Max, Team, and Enterprise plans and transcription tokens don’t count against rate limits.

I’ve always believed in voice becoming a first-class citizen interaction model. Building the first version of WhatsApp for the Meta RayBan Stories gave me a deep understanding of the nature and challenges of voice interaction.

On the one hand, it’s probably safe to say that most people are more comfortable speaking than writing – as children we speak before we write. For most, writing is a tool, whereas speaking is almost as natural as breathing. Speech, in a real sense, is closer to thought than written prose. Thus, voice interaction models arguably collapse some of the lag between thought and action.

On the other hand, I’ve always wondered what the physiological and social impacts of constant speaking might be. If we speak to our computers all day rather than type, what might the impact be on our bodies and minds? Can our modern, open plan offices handle it?

I, a Mavis Beacon trained millennial who uses Claude Code daily, look forward to venturing more fully into the world of voice. I look forward to seeing it’s impact not only on my coding, but also on my thinking.