Documentation
Voice typing
Speak instead of type, into a draft you read before anything reaches the command line.
For the AI agent that plans and runs steps, see the Harness guide.
Turn it on
Agree once, and sign in
Voice typing sends what you say to Termiyo’s server to be written out, so it asks you to agree the first time and needs the machine signed in to a Termiyo account. Nothing records until you have agreed.
A draft, not the command line
Your words land in a draft strip, not straight into the terminal. You read and edit it, then send it — Termiyo never runs anything from your voice on its own.
Speak
How it is handled
The microphone opens and the strip shows that it is listening. The recording is cut at its pauses into pieces and each piece is written out in order. If part of the recording did not make it into the draft, Termiyo tells you which part rather than leaving a quiet gap.
What is sent, and what is kept
Audio goes only to be transcribed
The page that records has no network of its own. The pieces are sent to be written out only after you have agreed and while you are signed in, through Termiyo’s server to a speech vendor. The server holds the vendor key and your AI allowance and keeps no audio.
Hindi comes back in Roman letters
Speech that transcribes as Hindi is written in Roman letters by default, so it does not break a terminal that cannot draw Devanagari. You can choose Devanagari instead if you want it.