shear
Upload raw. Broadcast-ready.
Shear transcribes your recording, finds every silence over 0.6s and every filler word, and removes them automatically. A tighter cut and baked-in captions in one command.
Replaces Descript · saves $30/mo
A 12-minute take becomes a clean 8-minute cut.
Word-level timestamps for every word, in 99 languages. Runs on your machine.
Every gap over 0.6s and every filler word is marked, with padding so no consonant gets clipped.
The clean cut renders with an SRT file generated and burned in.
So today I want to show you um the exact system that doubled like our email signups in 30 days. Basically it comes down to three things. You know timing, copy, and uh the offer structure.
Everything Descript does. None of the lock-in.
Word-level timestamps from our transcription engine. Runs fully locally, your audio never leaves your machine. 99 languages.
Configurable gap threshold (default 0.6s) with smooth padding so cuts never clip words.
Um, uh, like, basically, you know. All gone. Add your own with --filler.
SRT generated from the clean transcript and burned into the video. Style it with --force-style.
Nothing to install. Our processing engine ships with everything required. Mac, Windows, and Linux.
--transcript-out exports the transcript as JSON. Clip reads it directly, no second pass.
Questions, answered.
Anything else, write to hello@vertexcore.studio.
01How does silence detection work?
--silence-thresh (default 0.6s) is removed. A --silence-pad (default 0.12s) is kept around each word edge so consonants never get clipped.02What accuracy levels are available?
--model fast, medium, or high.03Can I customize filler words?
--filler "um,uh,like,so,right,honestly" to add words on top of the defaults: um, uh, hmm, like, basically, literally, you know, i mean, sort of, kind of.