Remove Filler Words From Dictation: Clean Voice Typing

Remove filler words from dictation on your Mac. Dictato strips um and uh by default and tidies stumbled repeats when you switch it on, all on-device.

You speak the way you think: in fits and starts. You start a sentence, lose the thread, back up, and start again. Then you say “um” three times while you hunt for the right word. A literal transcript keeps all of it, and then you spend the minutes you saved deleting your own “um”s and pulling apart the false starts. Dictato is built to hand you something more readable instead, so the page reads closer to a finished note than a recording of you thinking out loud.

What gets cleaned up automatically

Out of the box, with no setup at all, Dictato removes the hesitation sounds you make while thinking: the “um”, “uh”, “er”, and “hmm” that pepper natural speech. You do not flip a switch for this. You talk normally, and those filler sounds simply do not land in your text. That is the one bit of cleanup you get by default.

What you can switch on

The fuller cleanup is opt-in, and it is worth turning on. When you enable automatic correction (Auto-correct) or set up a per-app profile, Dictato also tidies the stumbled repeats and false-start restarts that come from talking out loud. Say “on the… on the second floor” and you get “on the second floor”. You caught yourself, relaunched the phrase, and the redundant run does not survive into the final text.

This same opt-in layer treats multi-word verbal tics like “you know” or “I mean” conservatively. It may trim them, but it does not chase down every one, and it will not touch them at all unless you have switched the cleanup on.

It keeps your words

The cleanup is deliberately cautious. It keeps your wording, your phrasing, and your meaning, and when a passage is already clean it leaves it alone. It is built to never invent content or change what you actually said, so you are not trading filler removal for the risk of words you never spoke. It handles redundant repeats and adjacent restarts, not full mid-sentence reversals, so do not expect it to untangle a “scratch that, let me start over”.

Tune it per app

Because the deeper cleanup lives in per-app profiles, you can set the level of polish app by app. Tidy your writing in one place, leave another fully verbatim where every word matters. If you want tighter control over the result, you can also drop in punctuation commands as you speak.

The honest payoff

By default you get your transcript with the hesitation sounds taken out. Turn on the deeper cleanup and you get readable text that still sounds like you, minus the stumbles you did not mean to keep. All of it runs on your Mac: no cloud, no account, nothing uploaded. You speak the messy way humans actually speak, and what lands on the page is closer to what you meant.