Turning voice on

Voice is off by default. Turning it on is a one-time setup: choose which engine you want, connect it with a key, and pick a voice. After that it’s ready whenever you want it.

Everything on this page lives in Settings → Hyo Claude Code Agent → Voice.

Choose an engine

The first setting is Voice engine. Pick GPT-Live for a real back-and-forth call, or ElevenLabs for the turn-taking version. What voice mode is explains the difference; if you’re not sure, start with GPT-Live.

Setting up GPT-Live

1. Get an OpenAI API key. Sign in at platform.openai.com and create a key under API keys. Calls are billed to this key at OpenAI’s per-minute rate.
2. Paste it into Hyo, into OpenAI API key.
3. Pick a voice. There are twenty-two to choose from. Press Play sample next to the picker to hear the one you’ve selected before you commit to it.

That’s the whole setup. Two more things on the same page are optional:

• Personality. One text box that says how the voice should carry itself on a call. It comes filled in with a sensible default (warm, direct, talks to you like a colleague who knows you). Change it if you’d like it drier, warmer, or more like you. If you mention your name here, the voice will use it when it greets you.
• Who it is. If you use your own agents, the voice introduces itself as whichever agent the tab is set to, using the name and description from that agent’s file. If you don’t have agent files, it’s simply your agent. There’s nothing to set.

Setting up ElevenLabs

1. Get an ElevenLabs API key. Sign in at elevenlabs.io and copy your key from Settings → API Keys.
2. Paste it into Hyo, into ElevenLabs API key.
3. Pick a voice. Once your key’s in, the Voice dropdown fills with the voices from your ElevenLabs library. Choose the one you like.

Two more settings for ElevenLabs, both optional:

• Playback speed: how fast it reads replies aloud. Starts at 1.25×; you can set it from 1× up to 2×.
• Auto-speak responses: on by default, so replies are read aloud automatically. Turn it off if you’d rather trigger playback yourself.

First time takes a moment

The first time you open voice mode on ElevenLabs, Hyo downloads two small models that run on your own machine. They’re what tell when you’re speaking and when you’ve finished a thought. It’s a one-time download, and after that voice opens instantly.

Turning it on in a conversation

With setup done, click the microphone in the status bar of any conversation to switch that tab into voice. The next page covers what you’ll see.

Switching engines later

Change Voice engine any time. Your keys stay saved for both, so switching back is one click and nothing needs pasting again.