Pauseless: A Free Browser Tool That Cuts Pauses and Ums From Video, Then Hands Your Editor the Timeline
Every talking-head video has the same hidden tax. A one-hour interview holds a surprising amount of nothing: breaths, thinking pauses, a guest searching for a word, an "um" before every answer. Somebody has to find each one and cut it, and that somebody is usually the person who would rather be shaping the story.
Tools that remove dead air have existed for years, and most of them share one assumption that turns out to matter a lot. They hand you back a finished video. The cuts are baked in. If the tool chopped the half-second pause that made a joke land, you are back to the original file, re-exporting, starting over.
Pauseless, a free tool that runs in a browser tab, makes a different bet. It finds the pauses and filler sounds, lets you review them, and then gives you a timeline for Premiere Pro, DaVinci Resolve, Final Cut Pro or CapCut where every cut is still just a cut. You can drag any of them open again. That one design choice is why it deserves ten minutes of your afternoon.
What it actually does
Pauseless showed up on Hacker News on November 4. Its maker, who posts as tahaygun, describes it as a "free silence/filler-word remover from videos, on-browser," and the project page is blunt about the business model: no account, no watermark, no time limit, MIT licensed. The code is on GitHub, and the repository went up at the end of October, so this is a very young project.
You open the page in Chrome or Edge on Windows or macOS and drop in a file: MP4, MOV, M4A or WAV. A few seconds later you see your recording laid out as a timeline with the quiet stretches marked for removal.
Four settings control what counts as dead air:
- Silence threshold. How quiet the audio has to be before it counts as a pause. There is an Auto option.
- Minimum silence length. Short gaps between words stay, longer ones go.
- Padding. How much breathing room to leave on either side of speech, so words do not start or end abruptly.
- Minimum clip length. Throws out tiny islands of sound like a cough or a mouse click that would otherwise survive as a two-frame clip.
Change a setting and the cuts update almost instantly, because the tool has already read the audio once.
Then there is a separate switch for filler sounds, "um," "uh," "hmm," with a sensitivity slider running from Fewer to More. Click any marked cut to keep it. Drag across a stretch of the timeline to mark a filler the detector missed.
When you are happy, you export. This is the part that matters.
The export is the product
Pauseless writes four kinds of timeline, and it can also render a finished video:
.fcpxmlfor Final Cut Pro and DaVinci Resolve.xmlfor Premiere Pro- a CapCut project, packed as a zip
.edl, the old plain-text edit list format most other editors can read- an
.mp4with the cuts applied, if you just want the file
The timeline files do not contain any video. They are instructions: "play the source from here to here, then from here to here." Your editor reads them and builds a sequence that points back at your original footage. There is an optional field for the folder where your video lives, so the editor finds the file without asking you to relink it.
That is why the export matters more than the detection. Automatic pause removal is never right everywhere. A comedian's pause before the punchline, an emotional beat where a guest collects themselves, the silence after a hard question: these are the moments a good editor protects, and they look exactly like dead air to a loudness meter.
With a baked video, every one of those mistakes costs you a re-export. With a timeline, a mistake costs you one drag of a clip edge. You keep the speed of automation and the final say of a human, and the tool never gets the last word on your story.
The rendered MP4 is there if you want it. It re-encodes the picture (using your computer's hardware video encoder when it has one), so expect a wait on long recordings, while the audio is copied over untouched. The timelines are ready the moment you click.
How it stays private and fast
Pauseless has no server. The project page says it plainly: "Your video is read from your disk, processed on your computer and never sent to a server." The maker's post on Hacker News adds that it "works completely on browser and has zero backend api connections."
This matters for more than privacy. Uploading an hour of 4K interview footage to a web service can take longer than the editing you were trying to save. Pauseless only reads the audio track, which is tiny next to the picture, and it does the work on your machine.
The speed figures come from the maker, so treat them as a guide: the project page lists a one-hour, 2.7 GB MP4 analysed in about 5 seconds in Chrome, the README says about 6, and the Hacker News post says roughly 10 seconds for two hours of video on a MacBook Air. Whatever your exact number, it is fast enough that you will spend far longer reviewing cuts than waiting for them.
The filler detection is the only part that uses anything you could call AI, and it is small on purpose. The README describes a tiny listening model, about 22,000 adjustable settings in total, trained on a public collection of recorded English meetings and fine-tuned on 32 minutes of hand-corrected Turkish podcast audio. It runs on your computer like the rest of the tool. You can also install the page as an app, after which it works offline.
Put this into practice
Here is a twenty-minute test that will tell you whether Pauseless belongs in your workflow. Use a recording you have already edited by hand, so you have something to compare against.
- Pick the right clip. Choose ten to fifteen minutes of a single person talking, ideally with one or two moments where the pause carried meaning.
- Start on Auto. Load the file, leave the silence threshold on Auto, and set padding a little generous to start. Tight padding is where jump cuts start to feel breathless.
- Turn on filler sounds at default sensitivity. Watch what it marks for the first two minutes before you trust it with the rest.
- Protect the good pauses. Scrub to the moments you know matter and click those cuts to keep them.
- Export the timeline, not the MP4. Choose the format for your editor and fill in the video folder path.
- Open it next to your hand edit. Play the two versions back to back. Count how many cuts you would change. If it is a handful, you just saved most of an hour.
Two habits help once you adopt it. First, run it on the raw recording before you start shaping the story, so the tedious cleanup is out of the way and your attention goes to structure. Second, keep a note of the settings that work for each show or speaker. A fast talker and a slow, thoughtful guest need very different minimum pause lengths.
If the filler detector misses words in your language, the export menu also offers a small .json file of filler labels, which the project page says you can share to help improve detection.
Where it falls short
The filler detector misses a lot. The README is honest about this: at default sensitivity it "finds about 70% of filler sounds with about 70-80% precision" on held-out English and Turkish speech. In plain terms, roughly three in ten "um"s will survive, and some of what it flags will be real words. Plan to listen through and fix by hand. Pause detection, which is just loudness, will be more reliable than filler detection.
Only Chrome and Edge, only Windows and macOS. It depends on video features built into those browsers. Safari and Firefox are not supported yet, and Linux is not listed.
Format limits. Audio inside your files needs to be AAC, Opus or plain WAV, and rendering an MP4 needs H.264 or HEVC video. Most camera and phone files qualify. Some professional formats, and recordings with unusual audio, will not.
It cuts on silence, not on meaning. It cannot tell a stumble from a deliberate restart, or a false start from a sentence. It will not remove "you know" or "like," and it will not pick the better of two takes. Those are still editor's decisions.
Your editor may already do part of this. Premiere Pro's transcript-based editing can find and delete pauses, and filler detection arrived there in beta in late 2023, according to a RedShark News walkthrough that also notes it was not fully reliable at catching fillers. If you live in Premiere and already transcribe everything, compare the two before switching. Pauseless earns its place for people on Resolve, Final Cut or CapCut, for anyone who does not want to transcribe first, and for anyone who cannot upload client footage.
It is days old. One person built it, it had a single star on GitHub when this was written, and nobody outside the maker has published tests. The MIT license means you can keep a copy running yourself even if it stops being updated, and the README shows how to build a single self-contained HTML file for exactly that. Still, expect rough edges.
Why this is worth your time
The interesting thing about Pauseless is not that it removes silence. Plenty of tools do that. It is that it treats the result as a suggestion rather than a verdict. Every automated editing tool should work this way: do the boring pass fast, then hand back decisions a human can change in the software they already know.
Try it on one episode. If the timeline it gives you needs a dozen fixes instead of four hundred cuts, you have your answer, and you have the hour back for the part of editing that actually needs you.