Skip to content

Voice input

How it works

Your speech becomes text you can edit before sending. Tap the microphone, talk, and the words appear in the message box as you go. Nothing goes to the AI until you tap send, so you can fix, cut, or rewrite anything first.

The AI does not speak back, and the microphone only runs when you have asked for it.

Starting from the home screen

If you add the app to your home screen, its Voice shortcut starts a new conversation with the microphone already listening, so you can start talking straight away. It is the only place the microphone starts without a tap. A voice link you type or someone sends you opens the same screen and waits for you to tap.

How long a session lasts

A listening session ends by itself after about two minutes, or sooner if you stop speaking for about twenty seconds. One tap starts another. The cut-off keeps a forgotten microphone from running. Voice input also has a daily allowance.

Where your audio goes

While you speak, your audio streams to GreenPT, our European AI provider, which turns it into text and sends the words straight back. GreenPT says audio is processed on its European servers, is not retained, and is never used to train models. We do not keep a copy either: the audio passes through our server on its way and is not saved. In the app you will see the short version at the moment you decide whether to speak: transcribed in the EU, audio not kept.

The energy for the AI's reply is counted as usual. The speech-to-text step is not counted: GreenPT does not report an energy figure for speech, and we would rather leave the gap than invent a number. See what we do not estimate.

Where you can't use it

Speaking sends audio to GreenPT, so temporary chats have no microphone. Type there, or start an ordinary conversation to speak.