Kannada Voice Typing

Kannada writes consonant clusters by hanging a subscript form below the preceding letter — the ottakshara. In ಕ್ರ the ರ sits underneath rather than beside.

For dictation this matters because there is no separate sound to recognise. The recogniser is not hearing an ottakshara; it is deciding which word you said and then writing that word’s spelling. When it picks the wrong word, the ottakshara is where the mistake surfaces — a subscript that should not be there, or one missing. It is small on screen and easy to read straight past, so it rewards a deliberate look.

ಎ against ಏ, and ಒ against ಓ, are separate vowels that change what a word means. Ordinary speech shortens the long ones, and the recogniser writes what it heard.

Together with the ottakshara this accounts for most of what you will fix. Both reward a targeted scan rather than a full reread, because the errors concentrate rather than spread.

The Kannada people speak and the Kannada people write are different registers. Dictation transcribes speech, so what you get is closer to conversation than to an essay, an application or anything official.

That makes it excellent for messages and notes, and a first draft for everything else. It is not a flaw in the recognition — it is doing exactly what it should. It just means the formal version is a rewrite rather than a correction.

The microphone is your browser’s. On Chrome and Edge the audio is sent to the browser vendor’s speech service and text comes back; WorldTyping is never in that path and never receives audio. The text stays in your browser, saved locally so a reload does not lose your draft.

That is also why dictation needs an internet connection, and why a browser without Kannada speech support will not respond to the button.