Answer your email by voice. Keep the words on the Mac.

Last updated 7 September 2026

The workflow, in any mail client

VV does not know it is in a mail app. It sees a text box with a cursor in it and types there, which is why the steps are the same in Apple Mail, Gmail in Safari or Chrome, and Outlook for Mac.

1. Open the message and click into the reply box, or start a new message and click into the body.
2. Hold the VV shortcut. The overlay shows a level meter so you know it is listening.
3. Say the reply. Speak the punctuation you want if you like, or let the model place it.
4. Release. The text is typed at the cursor and your clipboard is put back as it was.
5. Read it back once. Fix anything with the keyboard, exactly as you would a typed draft.
6. Send.

For a longer reply, double-tap the shortcut to latch it and talk hands-free, then tap again to stop. The subject line and the To field are ordinary text boxes too, so you can dictate into those, though most people find it quicker to type an address than say one.

Apple Mail

Reply, click into the body under the quoted message, hold, speak, release. Mail's own autocorrect still runs on top of what VV types, so a name it does not know may get "corrected"; add the name to Vocabulary and the problem goes away at the source.

Gmail in the browser

Click into the compose or reply area, hold, speak, release. VV types through the system input layer, so Gmail treats the text as if you had typed it: Smart Compose suggestions and Gmail's spell check behave as normal. It works in any browser because there is no extension to install.

Outlook for Mac

Same again: click into the body, hold, speak, release. If your organisation uses signatures set by an administrator, they stay put; VV only types where the cursor is.

What VV is not doing

There is no email integration to configure, and none coming that would read your mailbox. VV does not read incoming messages, does not see who a reply is to, and cannot send anything. It types words into a box. That is the whole surface, and it is deliberately small.

Editing tips that make dictated email read like typed email

• Dictate in short paragraphs. One thought per hold. A three-sentence reply is one hold; a long one is three or four, with a "new paragraph" between them. Short holds are easier to read back and easier to fix.
• Put your sign-off in a Snippet. A spoken trigger such as "sign off" can expand to your name, title and phone number. Do the same for your postal address, a booking link, or the two paragraphs you send every time someone asks for your rates.
• Teach names in Vocabulary. The people you email most are the words a general speech model is least likely to spell correctly. Add each one once with the spoken form, and VV biases towards the right spelling from then on.
• Read it back before you send. Dictation is fast; the read-back is where it becomes accurate. Ten seconds of reading is still faster than typing the reply.
• Optionally, run a Transform. If you want a formal register or a tighter version, a Transform rewrites the dictation on the way through. By default that runs on Apple Intelligence or a local Ollama model, so it stays on the Mac.

Before and after

What you said, holding the shortcut for about twelve seconds:

hi Priya um thanks for sending the deck over comma I've had a look and I think the pricing slide needs another pass full stop can we er can we grab twenty minutes on Thursday to go through it question mark new paragraph sign off

What you send, after a ten-second read-back and one deleted word:

Hi Priya,

Thanks for sending the deck over. I've had a look and I think the pricing slide needs another pass. Can we grab twenty minutes on Thursday to go through it?

Best,
Your Name
Head of Partnerships · 020 7946 0000

"Priya" came out spelt correctly because it was in Vocabulary. "Sign off" expanded from a Snippet. The hesitations were the one thing worth a keystroke.

Limits and the privacy boundary

Standard dictation is recognised on your Apple Silicon Mac and the text is typed there. No audio and no text is uploaded. The first launch downloads the speech model, which is the only time standard dictation needs the internet.

VV keeps a local history of what you dictated, with the audio, under the app's data directory. That is useful when a reply went sideways and you want to hear what you actually said, but it means dictated email lives on the Mac for the retention period you set in Settings. Keep it short on a shared machine.

A Transform can be pointed at a cloud provider if you choose to configure one. When it is, the text of that dictation is sent to that provider, and only the text; the app labels those Transforms as cloud so you are never surprised. Nothing about the surrounding email is sent, because VV never sees it.

Limits worth knowing: VV needs an Apple Silicon Mac on macOS 14 Sonoma or later. It will not fix a reply you would not have sent typed, and it does not summarise the thread you are replying to. The judgement is still yours.

Compared with Apple Dictation

Apple's built-in dictation also works in Mail. The differences are Vocabulary with spoken forms, Snippets, a local history with audio, and a rewrite step that stays on the Mac by default. We set the two side by side on the VV vs Apple Dictation page.