Talking Out Loud

A chat can listen and answer aloud. You speak as if on the phone and get an answer in a voice.

The major services have this in their phone apps and desktop programs. You turn it on with a separate button in the conversation window.

Voice mode is built into the service's app. The app records your speech, sends it to the server, and plays back the answer. How sound becomes text and text becomes sound differs by service.

In Claude the service transcribes your speech into text. The model receives that text and answers in text, and a separate speech synthesis program reads the answer aloud.

In ChatGPT and Gemini the model takes the audio itself. The models behind those modes are trained to understand speech without an intermediate transcript and answer in voice directly.

Where it is convenient

On the move. Thinking a task through on the way to the office, dictating a draft email, going over the plan for the day.

When saying it is easier than typing it. A long description of a task is spoken faster than it is typed.

For language practice. A conversation in a foreign language with corrections costs less than a tutor and is available at any hour.

Where it gets in the way

Precise text cannot be edited by voice. It is easier to see the line on screen and replace a word in it.

An open-plan office gets in the way too: you are saying out loud what colleagues do not need to hear.

With numbers and names. By ear they get confused more often than when typed.

What happens to your voice

The recording goes to the server exactly like text. Talking out loud is the same conversation with the service, only entered differently.

The rule is the same as for text. Anything you cannot type into a personal chat you cannot say out loud either: company information, client names, internal company numbers.

Take the people around you into account as well. The answer plays aloud, and you are not the only one who hears it.