# Flag B-roll shot descriptions that match narration beats

Judges: narration sentence plus one candidate clip caption from a vision model. Questions: choice:match - direct/illustrative/mood/unrelated; score:fit - 1-5. Action: Code builds a ranked B-roll shortlist per sentence; direct matches auto-placed on a timeline track.

Grade: A — A clean typed decision \(yes/no, a pick from a list, or a level on a scale\) that is concrete, repeats routinely, and has meaningful value.. Score: 4.22.

Vertical: Agencies, media & creators.

Function: Creative & industry.

Status: Idea.

Volume: routine.

Value: meaningful.

Risk: moderate.

Evidence: null.

Flags: None.

- [Vertical](/verticals/agency_media_creator/)
- [Function](/functions/creative_industry/)
- [Pattern: moodboard image clustering by caption](/use-cases/pattern-moodboard-image-clustering-by-caption/)
- [Print proofing comment categorization](/use-cases/print-proofing-comment-categorization/)
- [Logo submission contest screening](/use-cases/logo-submission-contest-screening/)
- [A visual reference finder](/use-cases/a-visual-reference-finder/)
- [A walkable 3D street view, classified by Jev](/use-cases/a-walkable-3d-street-view-classified-by-jev/)
- [Nitpicky, zooming into AI images](/use-cases/nitpicky-zooming-into-ai-images/)