Remove filler words and silences free

By ShinobiTools Team · Last updated: July 2026

Every "um", every "uh", every four-second pause where nobody says anything: found automatically and cut out on the millisecond. Your recording gets shorter and tighter, and you never touched a timeline.

Transcribe audio now →

Where the minutes actually go

Unscripted speech is full of gaps. Thinking noises between sentences, the pause while someone finds a slide, the silence after a question. Individually they're a second or two; across an hour of talking they're minutes, and they're the difference between a recording people finish and one they close.

Cutting them by hand is the least interesting work in editing: find the gap, set two markers, ripple delete, repeat two hundred times.

Two things it removes

Filler sounds — um, uh, erm, mm, ah, eh, in whatever spelling the transcriber heard. Found in the transcript itself, so each one is cut on its own start and end time.
Long silences — stretches below −32 dB lasting longer than about 0.7 seconds. A quarter of a second of quiet is left standing on each side, so sentences don't slam into each other.

"Like" and "you know" are deliberately not on the list. They're usually part of the sentence, and cutting them leaves speech that reads fine on paper and sounds mangled out loud.

It edits, it doesn't speed up

Nothing is sped up, pitched or time-stretched. Both the audio and, for video, the picture are cut at the same points, so voices sound exactly like themselves and lips stay on the words. What comes back is simply the same recording minus the parts where nothing was said.

Do it in the same pass as the swearing

Filler words and silences sit next to the swear-word options in the same panel, so one upload can hand back a recording that is shorter and bleeped. You still get the normal transcript, SRT and VTT out of the same job — see audio to text.

Common questions

How much shorter does it get?

That depends entirely on the speaker. A tightly scripted read loses almost nothing; a nervous unscripted interview can lose several minutes an hour. The result tells you how many pieces were removed.

Will it cut in the middle of a sentence?

It only cuts filler sounds and stretches where nobody is speaking, and it leaves a quarter-second of quiet on each side of a silence so the join doesn't sound abrupt.

Does it remove "like" and "you know"?

No, on purpose. Those usually carry meaning or grammar, and removing them tends to leave sentences that sound broken.

Does it work on video too?

Yes. Picture and sound are cut at the same points, so nothing goes out of sync. Video that is cut has to be re-encoded; beeping or muting alone copies the picture untouched.

Is it free?

Yes. 90 minutes and 2 GB per file, no daily cap, no account and no watermark.