Graphic tutorial: Turn Claude Code into a video studio with the OpenMontage skill pack, from loading to filming图文教程:用 OpenMontage 技能包把 Claude Code 变成视频制片厂,从装到出片
Then do it again: decompress the skill pack and → hang the 157 skill files into Claude Code. → Use a pipeline command to let the AI go through the complete process of scripting, framing, dubbing, and synthesizing.
跟着做一遍:解压技能包 → 把 157 个技能文件挂进 Claude Code → 用一条管线命令让 AI 走完脚本、分镜、配音、合成的完整流程。

Step 1: Prepare the Skill Pack第一步:准备技能包
Download this companion skill pack (zip only 0.64 MB) and unzip it to get the OpenMontage-skills directory with skill files, pipeline definitions, and access guides.
Read AGENT_GUIDE.md first. It is the official instruction for use written to the agent to understand how to call the pipeline and how to choose the tool.
For users who want a complete environment with sample material, press README to clone the official repository and execute pip install -r requirements.txt.
下载本文配套的技能精华包(zip 仅 0.64MB),解压后得到 OpenMontage-skills 目录,里面是技能文件、管线定义和接入指南。
先读一遍 AGENT_GUIDE.md,它是官方写给智能体的使用说明,了解管线怎么调用、工具怎么选。
想要完整环境(含示例素材)的用户,按 README 克隆官方仓库并执行 pip install -r requirements.txt。

Step 2: Hang skills into Claude Code第二步:把技能挂进 Claude Code
Copy the extracted skills directory as a whole to the user's home directory under .claude/skills/(Windows users create .claude/skills/in the C disk user directory).
Put the two directories of pipeline_defs and schemas into the project working directory, and the AI needs to read the pipeline definition when scheduling the task.
Restart the Claude Code session and enter "List your current OpenMontage skills". You can report scripting, storyboarding, dubbing and other skills to load successfully.
把解压出来的 skills 目录整体复制到用户主目录的 .claude/skills/ 下(Windows 用户在 C 盘用户目录建 .claude/skills/)。
把 pipeline_defs 和 schemas 两个目录放到项目工作目录,AI 在编排任务时需要读取管线定义。
重启 Claude Code 会话,输入「列出你当前掌握的 OpenMontage 技能」,能报出脚本、分镜、配音等技能即加载成功。

Step 3: Run the first pipeline第三步:跑第一条管线
Open a new project directory and say to Claude, "Use the OpenMontage pipeline to make a 30-second science short video with the theme of your choice. First, give me a storyboard."
The AI will define the first out script and split list according to the pipeline, confirm it, and then execute it step by step: generate commentary, match the material, and synthesize the soundtrack.
Plain text links (scripts, shots, subtitles) are used out of the box; links involving screen generation need to be connected to external engines such as ComfyUI or Kling according to the instructions in skills.
新开一个项目目录,对 Claude 说:「用 OpenMontage 管线做一条 30 秒的科普短视频,主题自选,先给我分镜脚本」。
AI 会按管线定义先出脚本与分镜表,确认后逐环节执行:生成解说词、匹配素材、合成音轨。
纯文字环节(脚本、分镜、字幕)开箱即用;涉及画面生成的环节需要按 skills 里的指引接入 ComfyUI 或 Kling 等外部引擎。

Step 4: Customize your production process第四步:定制你的制片流程
Open PROMPT_GALLERY.md, which has an example of an official tune-up prompt, and change it to your brand tone to reuse it.
The pipeline definition is plain text, and you can add or delete links: for example, add a "sensitive word self-check" before dubbing, and the AI will do the same.
It is recommended to have the AI review once after each sheet: write the link of the jam shell back to the skill file for comments, the more easy it is to use.
打开 PROMPT_GALLERY.md,里面有官方调好的提示词样例,改成你的品牌口吻即可复用。
管线定义是纯文本,可以增删环节:比如在配音前加一道「敏感词自查」,AI 会照办。
建议每条成片后让 AI 复盘一次:把卡壳的环节写回技能文件备注,越用越顺手。

Use reminder使用提醒
This article is compiled from public sources and ships with the matching resource. Product features and pricing are subject to the official page. Resources are for learning and exchange only — please respect the original license.
本文整理自公开资料并附上配套资源;涉及产品的功能与价格以官方页面为准。资源仅供学习交流,请遵循来源许可。
想马上用起来?去「AI工具」栏挑一个直接下载,或在「AI教程」里跟着图文步骤做一遍。
去 AI工具 → 看 AI教程 →阅读与点赞数据保存在你的浏览器本地,欢迎留下你的想法。
今日正能量学一点,用一点;今天种下的种子,会长成明天的能力。去免费下载专区 →广告
评论 文明发言,让讨论更有价值