Why filler words take so long to cut by hand
Cutting ums in a normal video editor means listening through the whole recording, spotting each one on the waveform, and making two cuts around it. A filler is often a third of a second long, and a ten-minute talk can have dozens of them. When we edited one of our own 12-minute videos by hand, it had 37.
In Transcut, the fillers are already marked in the text when the transcript appears. You read instead of scrubbing.
How it works
- Drop in your video. Transcut transcribes it on your Mac, with the exact timing of every word. Most transcription tools quietly tidy up your speech and leave the ums out. Transcut asks for every um to be written down, because you can’t remove what isn’t in the text.
- Fillers are marked as you read. Um, uh, erm, hmm and their long spellings (“ummm”) get an orange mark. So do stutters: a short word said twice (“I, I went”) or a word you started and restarted (“th- the”).
- Open Review. Every filler is in one list, with the reason (a filler sound, or Stutter: “I” said twice) and how sure Transcut is.
- Listen before you decide. Each suggestion plays as it is now and as it would sound with the change.
- Accept one, or all. Press Accept all to remove every sure filler in one go. The video is cut as you accept, and one undo brings everything back.
“Maybe” words: you stay in charge
Some words are fillers only sometimes. “I was, like, tired” has a filler; “I like this camera” doesn’t. Transcut marks these words (like, so, actually, basically, literally, well, right, you know, I mean, kind of, sort of) as Maybe, and only when commas set them apart the way a spoken pause does.
Maybes are never removed in bulk. When there are any, the button reads Accept N sure and leaves the Maybes for you to hear and decide one by one.
When a filler runs straight into the next word with no gap, the card says “no pause around it: listen first”, so you know which cuts to check.
How accurate it is
We measured it on eight recordings whose transcripts we checked word by word, by ear. Of the 34 filler sounds Transcut suggested, all 34 were real fillers, and all 5 suggested stutters were real stutters. In total it found 34 of the 39 fillers in those recordings; the 5 it missed were ones the transcription never wrote down, mostly in scripted clips.
Remove a word everywhere
Some fillers are real words: “basically”, “actually”, your own favorite. Search for the word, and Delete all removes every one of them in one step. One undo brings them all back.
Clean cuts, not choppy ones
Cuts land in the quiet between words, so a removed “um” doesn’t leave half a breath behind. You can fine-tune any cut to a hundredth of a second, and the exported video sounds exactly like the preview. To tighten the gaps that are left, see removing silences and long pauses.
On the Free plan
Your first 3 exports include every feature, Review included. After that, Review still finds and plays every filler, and deleting words by hand in the transcript always works; accepting Review’s suggestions needs Pro. See pricing.