long-to-shorts

Installation
SKILL.md

Long to shorts

Two jobs: find the moment, then cut and frame it.

Workflow

  1. Find the moment. Read the transcript for one coherent idea with a payoff, at the requested length. Do not impose 30 to 60 s on every extraction. If the requested moment is not in the source, say so plainly and change nothing rather than cutting something else.
  2. Trim to it. cutRanges or trimClip so the clip starts on the strongest sentence, keeps the context needed to follow it, and ends right after the payoff, never trailing into the next topic.
  3. Reframe to vertical. setProjectSettings aspectRatio "9:16", then one reframeAuto with follow:"faces" per clip: it frames every shot on its measured face. Do not keyframe shots one by one. Check nobody is half out of shot.
  4. Captions 3 to 4 words a line, y about +0.2 H so they sit above the bottom 22% (captions skill).
  5. Hook. Written from what is said in this clip, never generic bait. addTitle role hook when a text hook serves the brief.
  6. Audio. Level both speakers so one is not twice as loud as the other (setAudio compress and volume).

Gotchas

  • Cut on sentence boundaries. A voice line cut through a word blocks the run from finishing.
  • reframeAuto follow:"faces" needs measured face boxes (semanticProfile.visualContext); without them, inspect the source first.
  • Never letterbox a landscape shot into 9:16.
Installs
3
Repository
vidmoat/skills
GitHub Stars
1
First Seen
Oct 29, 2026