qiaomu-cut skill
A video-director skill: from one sentence it builds a project with footage, storyboard, captions, motion, an ffmpeg render and quality checks
- 378 ★ on the repository
- Last commit
- MIT
Broad access
Why this level
- The setup command installs programs through Homebrew, apt and npm, and the scripts run ffmpeg and a browser
- Paid ListenHub and 33台词 operations spend account credits and require keys and authorization
- Movie clips and third-party material may be protected by copyright, and the user is responsible for checking usage
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/qiaomu-cutDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- The setup command installs programs through Homebrew, apt and npm, and the scripts run ffmpeg and a browser
- Paid ListenHub and 33台词 operations spend account credits and require keys and authorization
- Movie clips and third-party material may be protected by copyright, and the user is responsible for checking usage
This is third-party code. Review the repository files before installing.
What it does
qiaomu-cut turns a request such as "make a 60-second English video about excavators" into a reproducible video project. The skill first builds an internal QiaoCut IR plan covering duration, format, audience, style, storyboard, sourcing strategy and renderer. Then it picks what is needed: free Pexels and Pixabay stock, local files, ListenHub services for images, voice-over and music, local Coli speech recognition, HTML, Manim and PPT. ffmpeg-full assembles the final video. Everything runs through node scripts/qcut.js commands. The setup command checks and installs missing dependencies: ffmpeg-full, fonts for Chinese captions, Playwright with Chromium, Manim, ListenHub CLI. There are separate workflows. Popular-science videos get Autopilot with seven scenes and a check of each. Motion graphics get the motion and audio commands, which fit the edit to a track's beat and master sound at minus 14 LUFS. Vertical explainers use explainer, and compilations of movie lines for English study use english-mix. Before paid actions the skill demands explicit confirmation with the --yes flag and a project binding, and it records downloaded material in a manifest with source and checksum. The finished video goes through one of three render tiers: preview, standard or final. The developer stresses that PPT export, complex masks and forced word alignment remain external capabilities and are not built into the skill.
Who it is for. For short-video makers, teachers and marketers who work in Claude Code or Codex on macOS or Linux and are ready to install ffmpeg and Node.js.
Good fit when
- You want a 60-second popular-science video: from plan and voice-over to a checked render with spot frames for every scene
- A recorded talking video should become a vertical clip with large captions, cutaways and beat-aligned transitions
- You need a promo with interface animation fitted to a music beat and sound mastered to minus 14 LUFS
- You want to compile movie lines for English study with checks on phrase boundaries and duplicate clips
- You want a quick draft in preview mode first and only then spend time on a final render
Not a fit when
- You want a single package with no extra programs: the skill requires ffmpeg-full, fonts, Chromium and, for some workflows, Manim and ListenHub CLI
- You work on Windows: the instructions mostly assume Homebrew on macOS and apt on Linux
- You cannot spend paid ListenHub or 33台词 credits: the skill creates such tasks only with an explicit --yes flag, but part of the features stay closed without them
- You need PPT output, speed ramps or full word-level caption alignment: the developer leaves these to external tools and does not build them in
- You plan to publish movie clips without rights to them: the user is responsible for checking usage terms of the material
Example request
Use qiaomu-cut and make a 60-second popular-science video on how attention works in transformers, with voice-over and spot frames for every sceneLimitations
- Node.js 18 or newer is required (20 or newer for ListenHub) along with ffmpeg-full. macOS with Homebrew is recommended, and on Linux dependencies install through apt.
- Movie clips need the 33台词 app and a separately installed, authorized 33tc adapter. The pick and cut commands create tasks that may spend account credits.
- ListenHub needs a key in the LISTENHUB_API_KEY variable or an OAuth login. The key should never go into commands, projects or the repository.
- The documentation is mostly in Chinese, and the default narrator voice, 向阳乔木 v1.1, belongs to the ListenHub service.
- PPT output, complex masks, speed ramps and forced word-level alignment are not part of the skill and remain external capabilities.
How to disable. Delete the qiaomu-cut skill folder from the agent's skills directory, for example ~/.agents/skills/qiaomu-cut. Dependencies that setup installed through Homebrew or npm have to be removed separately, while Playwright lives inside the skill's own .deps folder.
FAQ
How do I install qiaomu-cut in Claude Code?
Run npx skills add joeseesun/qiaomu-cut-skill --skill qiaomu-cut. For a manual install copy the qiaomu-cut folder into ~/.agents/skills. After that node scripts/qcut.js doctor --json shows what is missing, and the setup command installs missing dependencies.
What does qcut setup install?
It checks and, when needed, installs ffmpeg-full, fonts for Chinese captions, Playwright with Chromium, Manim and ListenHub CLI. Components already in place are skipped. macOS uses Homebrew, Linux uses apt and npm, and Playwright goes into a .deps folder inside the skill.
How do the render modes differ?
The skill distinguishes preview for quick iteration, standard and final. The content is locked in preview first and only then the longer tiers run. Motion mode adds a draft profile at half resolution and a final profile with three motion-blur subframes.
Does the skill spend money without my knowledge?
No. Creating ListenHub tasks, parsing and paid music operations require the --yes flag and a project binding through --qcut-project. For 33台词 the pick and cut commands create tasks that may charge credits, and they too are confirmed with --yes. Passing local files additionally needs --allow-upload.
How do I make a video from movie-line clips?
Use the english-mix fast path: init takes phrases separated by a bar, audit checks line boundaries, duplicates and streams, pace trims clips with pauses, and review builds spot frames for a phone check. An installed and authorized 33tc adapter is required.
Discussion
No questions or reviews yet.
Related
SkillsThe video-explainer skill turns a script into a vertical 9:16 explainer: narration, captions, Remotion and a pre-render review
A skill that turns a topic into a narrated 4K video: research, script, local TTS, Remotion composition and an FFmpeg mix
A skill that builds a narrated paper-collage explainer video from a topic, a talking-head clip or a single photo
Skill that turns a video into tall subtitle collages: crops burned-in subtitles or draws reviewed lines onto real frames
A skill that pulls highlights from a long video: download, transcript, segmentation, analysis and clips with titles
A Claude Code skill that turns a talking-head video into a short motion-graphics edit and checks its own cut
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| qiaomu-cut (this card) | no | 378 | |
| OpenMontage | no | 77,550 | a video studio with voice-over, images and stock, while qiaomu-cut stresses reproducible projects and render tiers |
| video-use | no | 30,591 | edits existing footage by cutting pauses and filler words |
| ffmpeg-skill | no | 4,619 | cuts and joins video with ffmpeg commands, with no storyboard or project |
README in short
The README first explains the idea: one request becomes a project with footage, storyboard, captions and checks. It then covers installation, dependency diagnostics, sample requests, ready workflows with commands, and the rules for ListenHub, voice-over and image generation.
README badge
Original SKILL.md
--- name: qiaomu-cut description: | 把一句话需求转成可复现、可验收视频工程的乔木智能剪辑导演。Use when the user asks to create, plan, edit, remix, explain, narrate, subtitle, animate, composite, or render a video—including one-line requests such as “制作一个科普视频：介绍 LLM 中的 RL”. Builds English-learning movie mixes, person profiles, explainers, code-drawn motion graphics (UI morph loops, showreels, kinetic-type promos, line-art/pixel/Three.js animation, lyric MVs), product launch videos, cinematic shorts, social clips, course/PPT videos, AI image/video/TTS/music-assisted stories, stock-footage stories, or multi-source montage. Routes across 33台词, ClipSeek/open media, local files, ListenHub, image generation, HTML/SVG/Canvas motion graphics with beat-grid storyboards and determinism probes, Manim, ffmpeg-full, captions, sound design, frame-0 video covers with YouTube thumbnails, and release checks. metadata: version: 0.16.0 author: 向阳乔木 copyright: Copyright (c) 向阳乔木 x: https://x.com/vista8 github: https://github.com/joeseesun/ mode: governed --- # Qiaomu Cut 你是乔木智能剪辑导演。目标不是“拼接文件”，而是把用户的一句话转成可验证的视频制作流程：理解意图、规划脚本、选择素材源、生成缺口素材、设计镜头语言、调用合适渲染引擎、质检并交付。 ## 触发条件 当用户要求做任何视频相关工作时触发，包括但不限于： - 电影/剧集台词混剪、英语学习视频、情绪表达片段合集。 - 人物介绍、品牌介绍、产品发布、课程讲解、知识科普、故事场景拼接。 - 免费素材搜索、AI 图片生成、B-roll 补齐、封面/片头/片尾制作。 - 为视频生成短镜头、旁白、多角色对白、音乐、解说源片，或从 URL 提取脚本材料。 - 转场、字幕、动效、遮罩、运镜、调色、音频响度、视频质检。 - 把网页、PPT、图表、数学动画、SVG/HTML 动效融入视频。 不要在这些场景触发：仅问 ffmpeg 基础命令、只要图片不做视频、只要搜索资料且没有视频输出意图。 ## 核心判断 先判断用户要的成品类型，再选工作流： 1. `english-mix`：影视台词 + 学习字幕 + 词卡 + 复读节奏。 2. `stock-story`：免费素材库 + 旁白 + 字幕 + 信息卡。 3. `person-profile`：人物资料 + 时间线 + 档案感包装。 4. `explainer`：Manim/HTML/SVG/PPT 风格解释动画。 5. `explainer-social`：竖屏概念动画 + 强字幕 + SFX + 社交短视频节奏。 6. `cinematic-short`：AI/素材图像 + 电影级运镜 + 音乐节奏。 7. `product-launch`：网页/产品 UI + Motion/HTML 动效 + 发布片。 8. `social-short`：竖屏短视频 + hook + 强字幕 + 卡点剪辑。 9. `talking-head`：口播精剪 + 字幕 + B-roll + 包装。 10. `data-story`：数据可视化 + 图表动画 + 旁白。 11. `motion-design`：代码逐帧绘制的动态图形——UI 形态变换、产品卡点片、Showreel、线稿/像素/Three.js 动画、歌词 MV。