Voice dictation

#How to dictate

1. Click into the field you want to write in.
2. Double-tap the trigger key. You'll hear a tick, and the thin line along the bottom of the box turns into a level meter — that's how you know it's listening.
3. Talk.
4. Tap once when you're done. Veda writes it up and puts it in the field.

Esc throws the recording away — nothing is written. Music and video go quiet while Veda listens and come back when you stop; switch that off with Mute other audio while listening in Settings → General. A recording stops on its own after two minutes and is written up as usual. If nothing was heard, Veda says so and leaves the field alone.

On a Mac, setup asks for the microphone; if you skipped it, macOS asks the first time you dictate. On Windows, desktop apps need microphone access switched on in Settings → Privacy & security → Microphone.

#What Veda cleans up

Speech is messy. Veda first works out the message you meant, then writes it the way you'd type it for the app you're in:

• Self-corrections — only the final version is kept. "Let's meet at 4, no wait, 5" becomes "Let's meet at 5." "Send it to Rahul, sorry, I mean Priya" becomes "Send it to Priya."
• "Scratch that", "delete that", "never mind" — dropped, along with whatever they cancel.
• Filler — um, uh, like, you know, basically, so yeah.
• Restarts and repeats — the clearest complete version is kept, once.
• Spoken formatting — "new line", "new paragraph", "comma", "full stop" or "period", "question mark", "bullet point". A spoken list ("first… second… also…") becomes bullet lines where that suits the app.
• Numbers, times and money are written normally: "two thirty" → 2:30, "fifty percent" → 50%.
• Names and terms from the conversation on screen are spelled exactly as they appear there, which fixes most mis-hearings.

#Telling it what to write

You don't have to dictate word for word:

• "Tell him I'll be ten minutes late" writes the message itself — not your instruction.
• "Reply yes and ask for the file" does what it says.
• With text already in the field, "make this shorter" applies to that text.

If there's already text in the field, what you dictate is added where it belongs — usually at the end. If you'd selected some text, what you say replaces the selection.

#Languages

Speak in any language. Veda keeps your language as spoken — Hinglish stays Hinglish.

#Where the audio goes

Nowhere. Your voice is turned into text on your own computer — Apple's speech model on a Mac, Whisper on Windows — and only the text goes on to the AI to be written up. Setup downloads the speech model once; you can check it in Settings → General, under Voice (Speech model). Without it, voice doesn't work at all — nothing is ever sent to a server instead. Home and the menu-bar or tray panel show Voice needs a download with a Download button, and if you double-tap or start meeting notes without it, Veda starts the download and tells you.

On a Mac, Apple decides which languages have a speech model. If your Mac's language doesn't have one yet, voice listens in English.

The recording is deleted as soon as it's turned into text. The microphone is on only between your double-tap and the tap that ends it.

#Turning it off

Switch off Double-tap to dictate in Settings → General, under Voice. A double-tap is then just an extra tap.