mcpField
V
video-frame-analysis

video-frame-analysis

github.com/mason0510Productivity371 installsUpdated 28w ago
Repository

About video-frame-analysis

video-frame-analysis is a lightweight skill for agents that extracts timestamped frames and OCR from local videos, then saves reviewable outputs and a text summary. It packages a repeatable evidence-extraction workflow rather than performing heavy video understanding.

Key features

  1. 01Reads video metadata with ffprobe
  2. 02Extracts frames at configurable intervals
  3. 03Builds a contact sheet
  4. 04Saves OCR text and JSON
  5. 05Writes a summary file

Use cases

  1. 01Review a local product demo recording
  2. 02Inspect a screen recording for visible text
  3. 03Collect evidence from a competitor walkthrough

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/mason0510/video-frame-analysis-skills --skill video-frame-analysis

After installing, type/video-frame-analysisin Claude Code to confirm it loaded.

Bundled files

References and scripts installed along with SKILL.md.

.gitignore
README.md
LICENSE
.github/workflows/
ci.yml
agents/
openai.yaml
references/
output-format.md
scripts/
video-frame-analysis.sh
tests/
test_video_frame_analysis.sh

Details

From the project README

video-frame-analysis

Overview

把“本地视频 → ffprobe → 带时间戳抽帧 → contact sheet → OCR → summary.txt / ocr.json 落盘”固化成一个可复用的轻量 skill。

适合给 Codex、Claude Code、OpenCode 这类 agent 处理本地 demo 视频、录屏、竞品页面 walkthrough。

When to Use

适用于：

  • 用户丢来一个本地视频，让你判断产品做到哪一步
  • 需要把视频拆成可复查的帧图、OCR 文本、摘要文件
  • 需要先做轻量证据提取，再做人工判断或进一步分析
  • 想避免每次手拼 ffmpeg / tesseract 命令

不适用于：

  • 网络视频下载
  • 批量视频队列
  • 实时流分析
  • GUI 或 Web 平台
  • 重型视频理解模型

Quick Start

bash scripts/video-frame-analysis.sh /path/to/demo.mp4 ./out/demo-analysis

默认输出：

  • ffprobe.txt
  • frame_*.jpg
  • contact.png
  • ocr.txt
  • ocr.json
  • summary.txt

Installation

推荐直接安装到 Codex / Claude Code / OpenCode：

npx skills add https://github.com/mason0510/video-frame-analysis-skills -g --all

安装后重启 agent 会话，再按需触发本 skill。

Input / Output

输入
  • 本地视频路径
  • 输出目录路径
输出
  • 视频元信息
  • 带时间戳帧图
  • contact sheet
  • OCR 文本
  • OCR JSON
  • 摘要文本

Config

FRAME_INTERVAL_SECONDS=5 FRAME_WIDTH=420 OCR_LANG=eng \
  bash scripts/video-frame-analysis.sh ./demo.mp4 ./out/demo
变量说明默认值
FRAME_INTERVAL_SECONDS抽帧时间间隔（秒）8
FRAME_WIDTH单帧宽度360
OCR_LANGOCR 语言候选chi_sim+eng

Project Files

  • scripts/video-frame-analysis.sh：主脚本
  • tests/test_video_frame_analysis.sh：最小闭环测试
  • references/output-format.md：输出目录说明
  • agents/openai.yaml：OpenAI/Codex 侧显示信息

Common Mistakes

  • 把它当成重型视频理解模型
  • 抽帧过密导致 OCR 噪声暴涨
  • 忽略 contact sheet，只盯 OCR 文本
  • 没检查本机是否装了 ffmpeg / tesseract

Comments

Markdown supported