clip
One long video. High-retention.
Clip scores every segment of your transcript for hook strength, selects the top three, and cuts them as vertical 9:16 clips with burned captions. Three hours of editing, compressed into 90 seconds.
Replaces Manual editing · saves 3 hrs/video
Hook scoring, in action.
Every transcript segment is scored 0 to 100 against six hook criteria.
The top three moments win, each with a written reason you can check.
Each clip is center-cropped to 1080x1920 with its own burned-in captions.
Strong curiosity gap and a practical framework reveal. Standalone arc with clear setup and payoff.
Direct actionable tip with specific numbers. Quotable phrasing. Ideal length for Reels.
Identity hook aimed at the audience's core aspiration. Tension resolved cleanly.
Three hours back. Every video.
Six criteria per segment: curiosity gap, practical value, emotional hook, standalone arc, quotability, and ideal length.
Detects source resolution and crops to a clean vertical. 1080x1920 at CRF 22, ready for Reels, TikTok, and Shorts.
Word-level SRT per clip, burned in. Every clip carries accurate captions without extra processing.
clips/manifest.json records rank, score, timing, and filename. Drop reads it to know what to publish.
Three clips by default. Pass --count for up to 10, each with its own score and reason.
Clip reads the transcript Shear already produced. One pass of speech recognition per video.
Questions, answered.
Anything else, write to hello@vertexcore.studio.
01How does hook scoring work?
02How many clips can I extract?
--count 5 (or any number up to 10) to extract more. Each clip gets its own score and reason in manifest.json.03Does it re-transcribe?
--transcript. Run standalone on raw video and it transcribes once, with your AI key set.