Free tool
Filler word remover
Take the filler words out of a transcript in your browser: paste the text, pick how careful the tool should be, and it marks every um, uh, you know and like before it deletes them. Nothing is uploaded.
Everything is counted in your browser. The text never leaves this page: nothing is uploaded, nothing is stored, no account is needed.
How to clean a transcript in four steps
-
Paste the transcript
Copy the text of the recording into the box. The tool reads it in your browser: nothing is uploaded, nothing is stored, and no account is needed.
-
Pick how careful it is
Three levels: the hesitations only, the hesitations with the discourse phrases and the hedges, or everything. A level switches the groups on, and you can switch any group off by hand.
-
Read the marks and the counter
Every filler word is marked in the text and counted by group. The counter tells you how many words the tool found and how big a share of the transcript they are.
-
Copy the clean text
The text without the filler words is written under the marked one. Copy it, or download it as a .txt file. The words, the punctuation and the order of the sentences stay as they were.
- Paste the raw transcript: the tool does not need the punctuation to work.
- Read the marked text before you take the clean one.
- The aggressive level removes words that can carry meaning — read it twice.
- Keep the original file: the tool never writes to it.
What the tool removes, and from which level
Four groups of words, three levels. A level switches on every group that is not more careful than itself, and you can switch a group off by hand at any time.
| Group | What it is | Removed from the level | Examples |
|---|---|---|---|
| Hesitations | The sounds a speaker makes while they look for the word. | Conservative | um, umm, uhm, uh, uhh, er, … |
| Discourse phrases | Phrases that hold a conversation together instead of saying something. | Standard | you know what, you know, i mean, sort of, kind of, more or less, … |
| Hedges | Words that add emphasis without adding information. | Standard | basically, actually, literally, obviously |
| Conversation fillers | Words that are often filler, but that can carry meaning. They are only removed at the aggressive level. | Aggressive | like, so, well, right, okay, alright, … |
Source: The list of the tool itself: apps/web/assets/tools/filler-word-remover/data/fillers.json, accessed 2026-10-22
What each level takes off the sample transcript
Questions this page answers
Does the tool send my transcript anywhere?
No. The counting happens in your browser, the page sends nothing, and the file never reaches a server. Close the tab and the text is gone; there is nothing to delete.
Which words does the tool count as filler words?
Four groups: the hesitations (um, uh, er, ah), the discourse phrases (you know, i mean), the hedges (basically, actually) and the conversation fillers (like, so, well, right).
What is the difference between the three levels?
Conservative takes the hesitations only, standard adds the discourse phrases and the hedges, aggressive takes every group. On the sample transcript of this page that is 6, 15 and 21 filler words out of 110.
Can I clean a transcript I got from Transcrib Si?
Yes. Transcribe the recording in the app, copy the text of the transcript and paste it here. The tool changes nothing else: the words, the punctuation and the order of the sentences stay as they were.
Does it work with a long transcript?
Yes. The text is read in the browser, so a two-hour transcript is no harder than a two-line one. A very long paste can feel slow on an old phone, and nothing is lost when it does.
Where to go next
The meeting as text, with the speakers told apart and a timecode on every line.
The recording as an SRT file, with the same transcript underneath.
One episode as text, ready to be read and cleaned.
Upload a recording and get the transcript in your own account.
Tools that work next to this one.
All of them run in the browser, on your own computer: nothing is uploaded and no account is needed. All free tools →
What Transcrib Si does, what it costs and where the file goes.
Check the reading speed of subtitles
Read an SRT or a VTT file and see which lines are too fast, too short or overlapping.
How long the cleaned text takes to say, at three measured paces of English speech.
Mark Burmistrov, Founder.
No transcript yet? Upload a recording and get the text with timecodes and speaker labels — then paste it here and take the filler words out.