Sentence-boundary cut
A sentence-boundary cut is an edit that begins and ends a clip at the start and end of complete spoken sentences, determined from word-level transcript timings, so the clip never opens or closes part-way through a word or phrase.
Most automatic clipping picks a start time and a length — sixty seconds from here — and cuts there regardless of what is being said. The result opens mid-thought and ends mid-word. Viewers leave within the first second, because a clip that starts in the middle of a sentence reads as broken before it reads as interesting.
How the boundary is found
It requires transcription with per-word timestamps, not per-block. Once every word has a start and end time, sentence ends can be located precisely, and the cut can be moved a fraction of a second to land on one. The clip is then defined by speech rather than by the clock.
- →Transcribe with word-level timings.
- →Detect sentence boundaries from punctuation and pause length.
- →Snap the intended start and end to the nearest boundary.
- →Accept a slightly different duration as the price of a clean cut.
Why the duration has to flex
A clip cannot be both exactly sixty seconds and cleanly bounded, because sentences do not end on round numbers. Something has to give, and it should be the duration. A fifty-four second clip that begins and ends on a complete thought outperforms a sixty-second one that does not.
What it changes for the viewer
The first second decides whether a clip is watched. Opening on a complete sentence gives the viewer a whole idea to catch hold of rather than the tail of one they missed. At the other end, a clean finish leaves the clip feeling deliberate instead of interrupted, which is what earns a rewatch or a share.
Every Clipd clip snaps to whole sentences by default.
See how it works