Dictate with voice input

Voice input turns what you say into text in the message box. Your words appear as you speak, ready to edit. Nothing is sent until you send it.

Dictate a message#

1. Press the microphone beside the send button, or press Ctrl+Shift+Space (Cmd+Shift+Space on a Mac). Your browser asks to use the microphone the first time.
2. Speak. A moving line shows that you are heard, and the words appear in the box. The words may change a little as you keep speaking and settle at each pause.
3. Press the microphone again, or the shortcut, to stop. Edit the text if you want to, then send it as usual.

Silence is not sent for transcription.

Choose the model and language#

Press and hold the microphone, or right-click it, to choose the speech model for this session. The default for your account is set under Settings, Chat, where you can also set the spoken language. With no language set, sven detects it.

Show words while you speak#

Under Settings, Chat, Show words while I am still speaking controls the live preview. With it on, the phrase you are speaking is transcribed about once a second, so you see words as you go, which costs a few times more than the spoken seconds. With it off, text appears after each pause and you are charged for the seconds you spoke.

What it costs#

Speech is charged by the second, at the price of the speech-to-text model you chose, shown on the pricing page per minute of audio. It appears in Usage and logs like any other charge.

Limits#

Speech is transcribed a phrase at a time, from recordings of up to five minutes. Streaming recognition over a single connection is not available. The formats are WAV, Ogg with Opus and WebM with Opus; the browser picks one for you.

From your own software#

The same speech-to-text models are available over the API. See Speech to text.