---
name: b-roll-gap-finder
description: Scan a video edit for long talking-head stretches without visual variety and suggest b-roll clips by topic.
title: B-Roll Gap Finder
category: image-video
difficulty: intermediate
author: admin
icon: 🎞️
input: mixed
output: markdown
phase: enhance
domain: content
tags: video-editing,talking-head,b-roll,scene-detection,transcript-alignment,llm-summarization,stock-footage,automated-analysis,content-production,editing-workflow,visual-variety,timestamp-detection
best_for:
  - YouTube and long-form video editing
  - Interview and podcast video post-production
  - Corporate training and explainer video creation
  - News and documentary editing workflows
---

## Description

Input: a rough-cut video + its transcript. Output: a list of timestamp ranges where the same camera angle persists for more than N seconds, plus a suggested b-roll topic for each range derived from the transcript at that point. Optionally matches the topic to a stock-footage library you provide.

## Why it works

The fastest way to improve a talking-head edit is to cut away during long single-angle stretches. Editors know this but miss it when they're deep in the edit. An automated scan surfaces the candidate gaps so the editor can fill them with intent.

## How it works

1. Scene-detect the video via pyscenedetect or similar — return cut points. 2. Identify stretches between cuts that exceed a threshold (default 20s). 3. Align each stretch with the transcript window at the same timestamp. 4. Summarize the transcript window into a 3-5 word topic hint via LLM. 5. Optionally, run the topic hint against a b-roll library's captions and return the top 3 matches. 6. Emit a markdown report with timestamp ranges, topic hints, and suggested clips.