What voice mode is

Voice mode lets you talk to your agent out loud, and it talks back. The way you’d talk to someone sitting next to you: you can interrupt, it can answer while you’re still thinking, and nobody presses anything.

Two ways to run it

Voice mode comes in two flavours. You pick one in settings and the rest of Hyo doesn’t care which.

GPT-Live is a real call. The voice listens while it talks, so you can cut in mid-sentence and it stops. It greets you when you pick up, chats while your agent is off doing something, and comes back the moment there’s an answer. This is the one to try first if you want voice to feel like a conversation rather than a tool. It runs on OpenAI’s live voice model, with your own OpenAI key.

ElevenLabs takes turns. You speak, it writes down what you said, your agent answers, and the reply is read aloud. It’s simpler, and if you already have an ElevenLabs account and voices you like, it’s ready to go. It runs with your own ElevenLabs key.

Both keep the conversation in the thread, both work on your phone, and both are per-tab, so one conversation can be voice while another stays text.

When it’s useful

Voice is good for the times typing gets in the way. Thinking a problem through out loud. Asking what’s on your plate while you make coffee. Getting an answer without sitting at the keyboard. Some things are just easier said than typed.

It’s off until you turn it on, because it needs a quick one-time setup. The next page covers that.

What you need to know first

• Voice runs through a separate service, OpenAI or ElevenLabs, which you connect with your own key. Calls are billed to that key, not to your Claude plan.
• Your microphone audio goes to that service so it can hear you, and its replies come back as audio. Your agent, your files and your Claude account work exactly as they do in text. The full picture is on Privacy & network use.