Dictate emails on your Mac — and actually sound like yourself

Most emails aren't hard to write — they're just slow to type. You know what you want to say, but between the opening line, the proofreading, and the typos, a five-line message eats five minutes. Voice solves the speed problem — you speak roughly three times faster than you type — but it creates a new one: a raw dictated email sounds casual, full of “ums,” and completely unpunctuated. The challenge when you want to dictate emails on Mac isn't just going faster — it's getting clean, properly toned text that's ready to send.

Why raw dictated email falls flat

Dictate a reply to a client using Mac's built-in dictation and look at the result: one unbroken block with no capitals or commas, your every hesitation transcribed word for word. That's fine for a personal note — not for a professional message.

Email has its own conventions, different from a text or a Slack message:

• correct punctuation — without it the text looks sloppy;
• the right register for the recipient — you wouldn't write to your manager the way you'd write to a close colleague;
• a minimal structure (one idea per paragraph, an opening and a closing line).

Apple's dictation does none of this: it's a transcription tool, not a writing tool. It faithfully reproduces what you said, hesitations and all. For email, you need an extra layer that turns a spoken stream into presentable written text.

The method: speak naturally, let AI handle the formatting

The approach that works flips the usual logic. Instead of dictating every comma and policing your language, you speak the way you would out loud, and an AI layer rewrites behind the scenes.

Concretely, a large language model (LLM) receives the raw transcript and cleans it up: it strips “ums” and repetitions, adds punctuation based on meaning, capitalizes correctly, and — crucially for email — adjusts the tone. You get a professional, direct message, not a verbatim transcript of what you mumbled.

An example

You hold your shortcut and say, in one go, inside your mail app:

uh hi so basically I just wanted to check if the quote was ready because the client is waiting on an answer today and uh ideally before noon thanks

Raw dictation gives you that block exactly as spoken. With AI formatting in “email” mode, the same stream becomes:

Hi,

Just following up on the quote — the client needs an answer today, ideally before noon. Could you let me know if it's ready?

Thanks in advance.

You didn't type a thing. The hesitations are gone, the punctuation is right, and the register has gone from loose spoken language to a clean professional email.

The right tone, chosen automatically based on the app

This is the difference that really matters day to day. Tone shouldn't require a manual adjustment with every message — it should follow wherever you're writing.

A good dictation app detects the active application and adapts the style accordingly:

The result: the same dictated sentence produces a polished email in your mail client and a short casual message in a chat — without you having to think about it. Your attention stays on what you're saying, not how it's wrapped.

How to make it frictionless

The time savings only hold if dictating is simpler than reaching for the keyboard. Three principles:

1. One gesture, in whichever app you're already in. The ideal is a global shortcut — for example, hold Ctrl, speak, release — that inserts text at the cursor, directly in the email field. No dedicated window, no copy-pasting.
2. Start with short replies. A few-line email is the highest-ROI dictation: you already know what to say, voice just accelerates it. Keep the keyboard for very formal or sensitive messages at first.
3. Proofread for the first week. While you're finding your rhythm, glance over the text before you send. You'll quickly notice that “proofreading” shrinks to a single look, and the habit takes hold.

What about your email privacy?

An email often contains sensitive information — client names, amounts, internal discussions. Two things to check before trusting an app with your dictations:

• Audio is not stored. Voice is used to transcribe, then discarded. Nothing should persist on a server.
• You control the AI provider. The BYOK (Bring Your Own Key) model lets you supply your own API keys (OpenAI, Gemini, Groq): your text travels through your account, under that provider's terms, with no opaque intermediary.

FAQ

Can I dictate directly into Mail or Outlook on Mac?
Yes. A dictation app that inserts text at the cursor works in any input field, including the body of an email in Mail, Spark, Outlook, or a webmail in the browser. Place your cursor in the message, dictate, and the text appears right there.

How do I stop my email from sounding “spoken”?
You need an AI layer that rewrites the transcript, not just transcribes it. The LLM strips hesitations, adds punctuation, and adapts the register to email format. You speak naturally; the result reads as written.

Do I need to dictate punctuation and line breaks?
No — that's the whole point. You speak at a normal pace; punctuation and structure are inferred from meaning by the AI. You never say “comma” or “new line.”

If your days are full of emails you know how to write but keep putting off typing, Speech Flow is built for exactly that: a native macOS app (Apple Silicon), lightweight (~50 MB), where you hold Ctrl, speak, release — and an LLM cleans, punctuates, and adapts the tone to your mail app, then inserts the text at your cursor. Privacy via BYOK (your own OpenAI, Gemini, or Groq keys, no audio stored), at €69 one-time, or an all-inclusive plan if you'd rather not manage any keys. The more email you write, the more obvious the payoff. To see how this approach compares to cloud-based solutions, check out the Speech Flow vs Wispr Flow comparison.