[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-96248":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":14,"subscribersCount":14,"size":14,"stars1d":14,"stars7d":14,"stars30d":14,"stars90d":14,"forks30d":14,"starsTrendScore":14,"compositeScore":15,"rankGlobal":9,"rankLanguage":9,"license":16,"archived":17,"fork":17,"defaultBranch":18,"hasWiki":19,"hasPages":17,"topics":20,"createdAt":9,"pushedAt":9,"updatedAt":27,"readmeContent":28,"aiSummary":29,"trendingCount":14,"starSnapshotCount":14,"syncStatus":30,"lastSyncTime":31,"discoverSource":32},96248,"reelbench-skills","eternityspring\u002Freelbench-skills","eternityspring","Learning notes and tooling skills for AI video - AI 视频相关的学习与工具 skill",null,"HTML",152,21,117,0,44.03,"Apache License 2.0",false,"main",true,[21,22,23,24,25,26],"ai-video","claude-code","claude-skills","ffmpeg","shot-analysis","video-analysis","2026-09-20 04:01:32","[![中文](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%E4%B8%AD%E6%96%87-285444?style=for-the-badge)](README.md)\n[![English](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEnglish-e2e6df?style=for-the-badge&labelColor=e2e6df&color=8b938a)](README.en.md)\n\n# reelbench-skills\n\n视频侧的 Claude Code \u002F Codex skill。\n\n| skill | 干什么 |\n| --- | --- |\n| [video-shots](skills\u002Fvideo-shots\u002F) | **拉片**：把一条成片拆成逐镜头的分析表——时长、景别、类别、运镜、画面。切点与时长由 ffmpeg 量，模型只判断该判断的四件事，14 道质量门逐条对账 |\n| [video-sync](skills\u002Fvideo-sync\u002F) | **合成带分镜信息的视频**：画面一边、分镜信息一边，镜头切了信息跟着切、镜头表自动滚动高亮。横版上下叠、竖版左右并，布局改一份 CSS 就行 |\n\n## 安装\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002Feternityspring\u002Freelbench-skills.git\ncd reelbench-skills\n.\u002Fscripts\u002Finstall.sh\n```\n\n软链到 `~\u002F.claude\u002Fskills\u002F` 和\u002F或 `~\u002F.codex\u002Fskills\u002F`（哪个装了就装到哪），**`git pull` 之后立刻生效**。\n\n```bash\n.\u002Fscripts\u002Finstall.sh --claude      # 只装到 Claude Code\n.\u002Fscripts\u002Finstall.sh --codex       # 只装到 codex\n.\u002Fscripts\u002Finstall.sh video-shots   # 只装某一个 skill\n.\u002Fscripts\u002Finstall.sh --uninstall   # 取消软链\n```\n\n依赖只有 `node` >= 18 和 `ffmpeg` \u002F `ffprobe`（macOS：`brew install node ffmpeg`）。\n**零 npm 依赖、零 API key**，用当前会话额度。\n\n不想软链就直接拷：`cp -r skills\u002Fvideo-shots ~\u002F.claude\u002Fskills\u002F`——skill 自包含，拷走就能用。\n\n## 示例\n\n`demo-report\u002F` 是拿 `demo-video.mp4`（202.9 秒的 AI 短片《啥是AI》）真跑出来的**完整产物**：\n53 镜、平均镜长 3.83 秒、每分钟 15.7 切、14 道质量门全绿。\n`demo-report-en\u002F` 是同一套流程跑一条 30 秒英文广告片的产物（`--lang en`，报告全英文）。\n\n[![拉片报告](skills\u002Fvideo-shots\u002Fassets\u002Freport.png)](demo-report\u002Fshots-report.html)\n\n报告是**单文件交互页**：内嵌播放器（播放时同步高亮镜头、点镜头跳转）、镜头节奏带、\n可搜索可筛选可排序的镜头表（列表 \u002F 卡片两种视图、首尾关键帧并排、点图开大图）、\n统计分布、出场人物、质量检查。零外部依赖，离线双击能开。\n\n\u003Cimg src=\"skills\u002Fvideo-shots\u002Fassets\u002Freport-mobile.png\" width=\"360\" alt=\"窄屏下的镜头表\">\n\n```\ndemo-report\u002F\n├── shots-report.html   ← 克隆下来双击就能开\n├── shots.json          ← 53 镜的拉片主数据\n├── shots.md            ← Markdown 镜头表\n├── track.json          ← 逐帧差分的运动曲线（机器证据）\n└── frames\u002F             ← 每镜首尾两张关键帧，共 106 张\n```\n\n`demo-sync\u002Fdemo-en-sync.mp4` 是 **video-sync** 的产出：同一条 30 秒英文广告片，\n画面在上、分镜信息在下，镜头切了信息跟着切。\n\n![带分镜信息的视频](skills\u002Fvideo-sync\u002Fassets\u002Foutput-landscape.png)\n","reelbench-skills 是一个面向 AI 视频分析与创作的轻量级工具集，专为 Claude Code \u002F Codex 等 AI 编程助手提供可插拔的视频处理技能。核心包含两大功能：video-shots（基于 ffmpeg 的自动拉片分析，输出带交互报告的镜头表，涵盖时长、景别、运镜等 14 项质量门检查）和 video-sync（生成画面与分镜信息同步显示的双轨视频）。项目纯前端 HTML 实现，零 npm 依赖、无需 API key，仅需 Node.js 18+ 和 ffmpeg 即可本地运行。适用于影视教育、AI 视频内容质检、短片创作复盘及教学演示等需要结构化视频语义分析的场景。",2,"2026-09-12 02:30:13","CREATED_QUERY"]