SuperAPI AI Directory

editor-camera-directing

Design & creation

View repository

Updated 10-18 04:44

About this skill

Editor Camera Directing

Requires

A running Scrollmark editor whose commands you can call: the SocialGPT editor chat, or scrollmark mcp from @scrollmark/cli for a coding agent. Without it the moves can't be applied; the rule for when a move is earned holds in any editor.

Directing, not a quota

This is directing, not a one-zoom quota and not a move on every clip.

Find beats in this order:

  1. Transcript — punchlines, reveals, "look at this," named objects, joke landings.
  2. The project's semanticBrief scenes (in the project snapshot) — silent B-roll, gestures, look-aways, props.
  3. media_inspectSourceFrames at 1–3 file seconds on those candidates. After apply, preview_inspectFrames at the timeline seconds.

A move only if the line is motivated and the subject (or a face, for a joke beat) is in the picture. If inspect cannot show it, skip. Several moves per video is fine. No count cap.

Applying a move

element_applyCameraMove on video/image only. Named kinds: zoom_in / zoom_out / pan_left / pan_right / tilt_up / tilt_down. Default amount 0.15. startS/endS are clip-local and cover the beat, not the whole take.

Stills/B-roll that would sit dead get a slow Ken Burns (zoom_in or a pan) over the hold. Talking-head: punch in on the beat; do not Ken Burns the entire speech. Match pan/tilt to where the subject is. Never element_setParams for motion (that clears camera keys). To take a move off a clip without undoing other edits, element_clearCameraMove.

Do not: a move on every keep-range; filmstrip the whole tape; animate captions/stickers.