references/a-roll

Decides which spoken content to keep or remove in A-roll dialogue editing.

38|4|Updated Jul 29, 2026
One-click install
npx skills add https://github.com/6174/recut --skill references-a-roll-6174
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: references/a-roll
Source: https://github.com/6174/recut/tree/main/service/skills/recut-director/references/a-roll
Command: npx skills add https://github.com/6174/recut --skill references-a-roll-6174

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Editing talking-head videos, interviews, podcasts, and courses requires deciding which spoken segments to keep, delete, or reorder. Without clear criteria, editors over-delete meaningful connective words, splice incompatible retakes, or strip natural breathing pauses, producing unnatural dialogue. ## Core Features & Use Cases - Semantic Unit Editing: Edits speech by complete sentences, answers, and ideas rather than keyword fragments, preserving connective words and context. - Filler Word Classification: Applies a three-category table to distinguish safe-to-delete fillers, context-dependent fillers, and functional connectives. - Retake and Pause Handling: Identifies true retakes, removes only fully covered failed attempts, and compresses long pauses to about 0.3 seconds without eliminating natural breath. - Use Case: Given a transcribed interview recording, use this Skill to clean filler words, remove failed retakes, tighten pauses, and restructure answers on an editable transcript before applying changes to the timeline. ## Quick Start Use the a-roll skill to clean up the transcribed talking-head footage by removing filler words and failed retakes while keeping complete semantic units.

Frequently Asked Questions about references/a-roll

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I decide which parts of an interview to cut?▼

Edit by complete semantic units such as full sentences, answers, or steps rather than keyword fragments. Keep connective words and context that retained sentences depend on, and when boundaries are uncertain, keep the content or make a smaller cut.

Which filler words can be safely removed from dialogue?▼

Only meaningless fillers like um, uh, er, and ah are safe for batch removal. Context-dependent fillers like so, like, or then require manual judgment, and functional connectives expressing causality, contrast, or sequence must always be kept.

How should retakes be handled in video editing?▼

First confirm the attempts are true retakes of the same intent, then keep the most complete version, usually the later one. Delete only the failed portions fully covered by the kept version, starting the cut at the failure point without removing useful earlier setup.

How much should pauses be shortened in spoken video?▼

Compress pauses longer than 0.8 to 1 second down to about 0.3 seconds, keeping 0.3 to 0.5 seconds between sentences for natural breathing. Never eliminate pauses entirely, and follow explicit user thresholds when provided.

What are the limitations of transcript-based dialogue editing?▼

Pause compression only works on gaps between transcript segments, not within-sentence breaths, which need word-level timing. The skill covers only speech keep-or-delete decisions; visual jump-cut coverage belongs to b-roll and overall pacing belongs to editing skills.