Skip to content
Back to skills

Xiaolu Motion

ASecurity

口播/录屏短视频剪辑与包装的导演工作流,配一套可复用的动效镜头库:确认选题 → 定风格 → 出分镜样帧 → 剪辑出片并过 QA 门禁,产出可直接发布的短视频成片。用于剪辑或包装口播/录屏类短视频(小红书、抖音、Shorts、Reels)时触发:删停顿、加动态字幕和运动图形、挑选视觉风格、或交付前跑像素级 QA 检查。Director workflow and reusable motion-graphics shot library for turning a talking-head or screen-recording video into a polished, packaged short-form edit — style selection, storyboard previews, effect shots, and a QA gate before final render. Use when editing or packaging a talking-head/voiceover video for short-form platforms (Xiaohon...

  • 5 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 7, 2026
designswiftapiperformance

Works with

  • claude code
  • api

Security analysis

A100/100

Pro scans all 20 files and shows the line behind each finding

Scanned November 7, 2026

npx -y skills add Wanfeng1028/muse --skill xiaolu-motion --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Xiaolu Motion?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Xiaolu Motion
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/wanfeng1028-xiaolu-motion/badge)](https://www.skillsdirectory.com/skills/wanfeng1028-xiaolu-motion)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: xiaolu-motion
description: "口播/录屏短视频剪辑与包装的导演工作流,配一套可复用的动效镜头库:确认选题 → 定风格 → 出分镜样帧 → 剪辑出片并过 QA 门禁,产出可直接发布的短视频成片。用于剪辑或包装口播/录屏类短视频(小红书、抖音、Shorts、Reels)时触发:删停顿、加动态字幕和运动图形、挑选视觉风格、或交付前跑像素级 QA 检查。Director workflow and reusable motion-graphics shot library for turning a talking-head or screen-recording video into a polished, packaged short-form edit — style selection, storyboard previews, effect shots, and a QA gate before final render. Use when editing or packaging a talking-head/voiceover video for short-form platforms (Xiaohongshu, Douyin, Shorts, Reels): cutting silence, adding kinetic captions and motion graphics, picking a visual style, or running a pixel-level QA pass before delivery."
---

# xiaolu-motion — 口播视频剪辑与包装 Skill

给任何创作者用的口播视频剪辑包装系统。输入逐字稿/分镜想法 + 真人口播或录屏素材,输出过 QA 门禁的短视频成片。核心思路:**代码即分镜**——JSON 分镜描述画面,浏览器(headless Chrome + WebGL2)渲染透明动效层,ffmpeg 分层合成到底片,Apple Vision 做像素级把关。不做可视化剪辑台,Agent 直接读文字、写分镜、出成片。

引擎、组件、渲染器、QA 门禁已是仓库一部分(`engine/` `components/` `render/` `qa/` `timeline/`);本文档讲的是**工作流和固定规则**,不是 API 文档——API 细节见各目录代码注释和 `README.md`。只在 macOS 完整可用(Apple Vision 做检测,`swiftc` 编译探针),环境要求见仓库 `README.md`。

## 四步流程

每一步做完都要拿给用户确认,得到明确点头才进下一步——风格和分镜这两步尤其不要抢跑,返工成本远高于多问一句。

1. **确认选题**:复述你理解的核心信息和目标平台/画幅,让用户确认。逐字稿没有词级时间码就先跑 ASR(思路见 `timeline/words.py`),分镜时间点可以用词锚 `{"text": "…"}` 代替绝对秒数。
2. **定风格**:挑 2-3 个候选(`styles/` 已有风格包,或 `library/` 演示成片截图),渲一张静态样张对比甩给用户选,选定后才进入分镜——风格决定后面所有令牌引用,返工要重写每一镜。用法见 `references/styles.md`。
3. **出分镜样帧**:按逐字稿写分镜 JSON(schema 见 `timeline/SCHEMA.md`),挑关键镜头渲静态样帧确认构图/字号/节奏,通过再渲全片——渲全片是四步里最贵的一步。参考做法见 `references/library.md`。
4. **剪辑出片**:先删停顿再决定是否提速(`references/editing.md`),渲**低清全片**确认剪辑点/字幕/节奏,再出高清成片,跑 QA 门禁(`references/qa.md`)——QA 不过就是没做完,不是"差不多能用",不把 `QA_FAILED.mp4` 当交付物。

## 固定规则

分镜怎么写、组件怎么调参都要服从这些设计前提,不是可选建议:

- **纸条字幕**:字幕是"贴在画面上的纸条"(见 `library/` 纸条字幕条目),不是无背景纯文字描边。
- **重点词放大配音效**:关键词短暂放大/加高亮底 + 轻 pop 音效(`components/kinetic_keyword.js`),不靠文字动效自己抓注意力。
- **人物不消失**:切进全屏图形/图表时真人收进右下 PiP,不整个消失;只有极短全屏转场例外。
- **不压脸、不压嘴**:图形/字幕/卡片不压脸框,尤其嘴部(脸框下三分之一)——QA 硬规则,写分镜时就该避开。
- **平台名/品牌名可配置**:分镜、组件参数里的平台名/产品名/品牌色做成参数或令牌,不硬编码。
- **无人段 ≤ 3 秒**:连续无真人画面(纯图形转场/素材插入)不超过 3 秒。

导演视角上"这一镜好不好"怎么判断(钩子、节奏、镜头设计、人物可读性、自检顺序)见 `references/director.md`。

## 需要时读哪份参考

| 何时读 | 文件 |
|---|---|
| 删停顿 / 整体提速的具体参数和命令 | `references/editing.md` |
| 并发内存锁、`.partial` 写法、ffmpeg 帧对齐、Vision 探测防泄漏、子帧运动模糊采样数 | `references/performance.md` |
| QA 门禁完整清单、先低清后高清的交付节奏 | `references/qa.md` |
| 怎么在 `library/` 镜头库里挑效果、移植参考实现 | `references/library.md` |
| 怎么写/继承一个风格包(`styles/*.json`) | `references/styles.md` |
| 判断一镜好不好:故事钩子、节奏、镜头/转场设计、人物可读性、自检顺序;反相混合 HUD 文字和粒子聚字两个具体技法 | `references/director.md` |
| 换成自己的固定开场素材(照片、封面) | `components/ip_intro/README.md` |
| 镜头库总表、预览视频 | `LIBRARY.md` |

## 兼容性

这套 SKILL 按 Agent Skills 通用格式编写,不绑定具体 Agent 产品——在 Claude Code、Codex 等支持读取仓库内 `SKILL.md` 的 Agent 里都可以直接引用。

Files in this skill

  • LIBRARY.md23.3 KB
  • LICENSE1.1 KB
  • README.md12.7 KB
  • SKILL.md4.9 KB
  • audio/README.md14 KB
  • audio/cli.py11.7 KB
  • audio/requirements.txt288 B
  • audio/sfx/CATALOG.md17.9 KB
  • audio/tests/test_xlaudio.py6.1 KB
  • audio/xlaudio/__init__.py1.5 KB
  • audio/xlaudio/analysis.py15.2 KB
  • audio/xlaudio/cues.py4.8 KB
  • audio/xlaudio/dsp.py22.9 KB
  • audio/xlaudio/grid.py5.5 KB
  • audio/xlaudio/ident.py8.9 KB
  • audio/xlaudio/instruments.py17.8 KB
  • audio/xlaudio/meter.py3 KB
  • audio/xlaudio/mix.py13.8 KB
  • audio/xlaudio/motion.py9.4 KB
  • audio/xlaudio/score.py24.7 KB

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.