Filler words

Choose whether to remove fillers

Remove filler words is on by default. It removes words from your saved list after transcription, without a rewrite model or network request.

The default list includes hesitation sounds such as um, uh, erm, and hmm. Words such as like, well, and so are not removed by default because they often carry meaning.

Edit the list

1. Open Settings → Transcription and find Filler words.
2. Turn Remove filler words on or off.
3. Type into Add a word, then choose Add or press Return. Click an existing word’s chip to remove it.
4. Use Reset to restore the default list.

Matching ignores capitalization and applies to whole words. Removing um will not remove those letters from the middle of a longer word. A comma or period immediately following a removed filler may be removed too, and the remaining spacing is tidied.

Check the list for your language

A sound that is a filler in one language can be a real word in another. For example, the default list includes er, which means “he” in German. Remove that entry if you want to keep it in German dictation.

The same list applies across your dictations, so keep the entries that suit the languages you use. You can turn filler removal off at any time.

How it fits with other cleanup

The order is filler removal → saved corrections → optional rewrite. Turning rewriting off or choosing No rewrite leaves filler removal and corrections active.

A cleanup rewrite profile can also remove false starts and adjust punctuation. That is a separate model-driven step; this list performs the specific removals you configure. For every setting involved in keeping the speech model’s result unchanged, see Raw transcription.