Guides··3 min read

Why "um" and "uh" ruin voice typing, and how AI removes them

Filler words are normal in speech and terrible in text. Here is why everyone uses them, why classic voice typing keeps them, and how AI dictation removes them without changing what you meant.

um, so, basically YOU SAY so um basically what I'mtrying to say is uh weneed more time AWRITE WRITES We need more time.

Record yourself explaining something for one minute and play it back. You will hear "um", "so", "basically", "like", a few restarted sentences and probably one "you know". You are not a bad speaker. You are a normal one.

Why we all use filler words

Speaking is real-time planning. While your mouth finishes one phrase, your brain is still choosing the next one. Fillers hold your turn in the conversation while you think. Listeners barely notice them. Researchers who study speech have found that filler words even help listeners follow what comes next.

Text is different. In writing, a filler is just noise, and it makes you look unsure or careless.

SPOKEN

So um basically what I'm trying to say is uh we need like two more days for the the design

WRITTEN

We need two more days for the design.

The kinds of filler AI dictation removes

TypeExamples
Soundsum, uh, er, hmm, aa
Filler wordslike, basically, actually, you know, I mean, so
Hindi and Hinglish fillersmatlab, woh, toh basically, haan toh
Repeats"the the", "I will I will"
False starts"Can you, could you send it"
Corrections"Monday, no wait, Tuesday"

The hard part: knowing what is filler

A simple "delete every um" rule would be easy. Words like "like", "so" and "actually" are harder because they are sometimes filler and sometimes meaning.

  • "It was like really good" → filler, removed.
  • "I like the second option" → meaning, kept.
  • "So the meeting moved" → filler, removed.
  • "It rained, so the meeting moved" → meaning, kept.

That is why this needs a language model rather than a word list. It reads the whole sentence and decides from context, the way a human editor would.

Does removing fillers change your voice?

It should not, and in Awrite it does not. We remove noise, not personality. Your words, your slang and your tone stay. A casual "haha ok fine, see you" does not become "Very well, I shall see you then."

Try this: next time you dictate, do not try to talk cleanly. Talk the way you would to a friend. The cleanup works best when you are not overthinking it.

What it looks like in a real message

Hey, I checked with the vendor. They can deliver on the 12th, but only if we confirm by tomorrow. Should I go ahead?1:18 ✓✓

What was actually said: "hey so um I checked with the uh vendor and they can like deliver on the 12th but only if we we confirm by tomorrow so should I uh go ahead". Twenty-six spoken words became twenty-two written ones, and every removed word was noise.

Will I stop saying um?

Probably not, and you do not need to. Some people notice they get more fluent after weeks of dictation, because they start thinking one sentence ahead. But the point of AI dictation is that it does not matter. You can talk the way you naturally talk and still send text that looks considered.

Quick answers

Why do I say um so much?

Everyone does. Filler words give the brain a moment to plan the next words. Most people use several per minute in normal speech, and it is not a sign of poor speaking.

Does AI dictation remove filler words automatically?

Yes. Awrite removes um, uh, repeated words and false starts automatically, while keeping words that carry meaning.

Will it remove words I actually meant to say?

No. Words like "like" or "actually" stay when they are part of the meaning, such as "I like it" or "it actually worked". Only filler use is removed.

Awrite Team
We build Awrite, AI voice typing for Android. Published 10 Nov 2026.

Keep reading