How to use Kontho AI

Read a short script aloud, train a voice on it, then type anything and hear it back in your own voice. Your recordings stay in your browser until you choose to train.

  1. 1

    Record the script

    Open Record and read the lines out loud, one take each. The script is built to cover the sounds a voice needs: statements, questions, exclamations, numbers, and long sentences that show your natural rhythm.

    Around 64 lines, roughly four and a half minutes of speech. Fewer works, but the voice gets less consistent.

  2. 2

    Keep the good takes

    Record a line as many times as you like. Press G to mark your favourite gold and S for a backup. Gold also becomes the active take, so marking is the only step — there is nothing else to confirm.

    Your gold takes are what the finished voice is built from, so favour the reading that sounds most like you at your most neutral — not the most expressive one.

  3. 3

    Trim the dead air

    Silence at the start or end of a take teaches the model to pause where you did not mean to. Open the editor on any take to trim it, split it into chunks, reorder them, or cut a chunk out. The editor detects the speech and shows you what it would trim; accept it or drag the handles yourself.

    Editing never overwrites. An edit is saved as a new take with the original kept beside it.

  4. 4

    Train your voice

    When you are happy with your takes, press Train. Your marked takes are sent for training and deleted afterwards — the trained voice is kept, the recordings it was made from are not.

    Around fifteen minutes. You can close the tab; it carries on without you and the voice is waiting when you come back.

  5. 5

    Speak with it

    Go to Voiceover, type or paste your text, and press generate. Clips are saved automatically and listed beside the editor, where you can replay, download, star or delete them.

    Press / to search the history once it fills up.