Set up dictation that keeps up.

It's easy to try dictation once, fight it for a few minutes, and go back to typing — because it felt slow, or it mangled a name, or a paste vanished and took the sentence with it. A lot of that isn't inherent to talking instead of typing; it's setup. Here's how to get Phrasora to the point where it feels faster than your keyboard, in a few minutes.

Pick a trigger you won't think about.

The whole point is that dictation should start the instant you decide to talk, not after you reach for a menu. On the Mac and Windows apps you hold a key, speak, and let go — the text arrives where your cursor already is. The default works, but the trigger is yours to change: every shortcut is editable, including a hold-to-talk key like Globe or right Command.

Choose something your hand already rests near. If you have to look for it, you'll stop using it.

Talk like a person, not a robot.

You don't need to enunciate, slow down, or say "comma" and "period" out loud. Phrasora transcribes with a full-accuracy speech model, then cleans the result with AI — punctuation, capitalization, capital letters, and filler removed. The "um," the false start, the "no wait, Friday" — all handled. On 73 real-speech recordings the median word error rate before cleanup was 2.33%, and the whole round trip, from key release to text at your cursor, runs a median of 871 ms.

So the instruction is: stop performing. Say the sentence the way you'd say it to a colleague.

Choose the formatting mode for the job.

The same spoken words should land differently in an email than in a code comment. Phrasora has three modes:

• Clean — lightly tidied prose. The default, and right most of the time.
• Polished — grammar and flow smoothed, still prose, never lists or headings.
• Structured — proactively shaped with headings and bullets when you're enumerating points.

On the Mac, Windows, and the apps you can set this once. On the iPhone keyboard there's a format mode in the top bar — Auto, Email, Chat, or Note — that resets each time so a choice in one app can't bleed into the next. Pick the mode that matches where you write most, and let Auto handle the rest.

Teach it the words that matter to you.

Every job has its own vocabulary — client names, a product, a framework, the way you spell your own surname. You can add terms by hand, but the more useful path is to just dictate and correct. When you fix a misheard term in your text, Phrasora notices and learns it, and the correction follows you to every device you're signed in on.

It announces what it learns, the moment it learns it, with a few seconds to undo — so you're never surprised by a term you didn't mean to add. Spell a name out loud once — "Sandhu, S-A-N-D-H-U" — and it locks the spelling in, without duplicating the word.

Set it up so nothing gets lost.

This is the part other tools skip. If focus moves while you speak, or a paste lands in the wrong place, or you dictate into a field that won't accept text, your words are captured to the Slate automatically instead of disappearing. There's nothing to configure — it's on by default. Knowing the worst case is covered is what lets you dictate quickly without watching the screen.

That's the whole setup.

Pick a trigger, talk normally, choose a mode, correct as you go, and trust that nothing is lost. Phrasora is free for 2,000 words a week, no card required — enough to find out in a few sessions whether it keeps up with you. See platform availability for the app on your device, or read exactly how your audio is handled before you start.