Talk. It types. English, Hindi, or Hinglish — and your voice never leaves this tab.
Hold Alt, say what you mean, let go. The speech model runs inside your browser, so there is no upload, no account, no word limit and nothing to pay for.
- Runs on your device
- No word limit
- No account
- Works offline after first load
Why this one is free and the others are not
Every dictation tool charges by the word because every word costs them GPU time on a server they rent. Wispr Flow gives you 2,000 words a week and then asks for $15 a month. Spoke has no server to rent: the model runs on the machine you are already holding, so the marginal cost of your ten-thousandth word is the same as your first. Nothing.
Your voice stays here
The audio is never uploaded, because there is nowhere to upload it to. That also means the IT policy that blocked your last dictation tool has nothing to object to.
Hinglish, written how people write it
Not kyā kar rahē ho. kya kar rahe ho — with "meeting" and "update" left in English, the way you would actually type them.
It works on a plane
After the first load the model is cached on your device. No connection needed, ever again.
What it does not do yet
Hindi mode forces Devanagari; it does not translate English into Hindi. Whisper can translate any language into English but never out of it, so English-to-Hindi needs a second model that is not in this build. Speak Hindi and Hindi mode writes it properly. Speak English and it will do its best, which is not the same thing — and saying so here is cheaper than letting you find out in a document.