开源项目解析
video-shotcraft
为 Claude Code / Codex 提供产品宣传片镜头配方,以网页截图与文案为素材,用 Remotion 渲染;当前 README 的 Gallery 列出 150 个镜头。
许可证附加条件:音频素材另按各自许可,见 ATTRIBUTION.md
读懂层
2 张 · 架构图 / 流程图
DeepWiki 架构页 · 抓取 2026-08-16 · 核验 2026-08-16
DeepWiki 架构页 · 抓取 2026-08-16 · 核验 2026-08-16
Agent 读 SKILL.md → 从 `references/shots/*.md` 里挑镜头 → `capture-template.mjs` 把产品网页截成贴图与 `layout.json` → `PageCam` 按 layout 做镜头运动 → Remotion 渲出 mp4。自带一条验证过的模板:36.2 秒、1920×1080、30fps、10 个镜头。
它管哪几个决定
自持能力
不涉及生成模型;渲染确定,随机量由代码里的种子函数控制
动手层
装它要过哪几关
- 1让 Agent 装README 的装法是让 Agent 自己 clone 并链进 skills 目录 —— 不是手动 pip/npm
- 2备 Node 与 Chrome截图走无头浏览器,少了 Chrome 拿不到贴图
- 3跑模板先照 36.2 秒那条模板跑通,再改镜头 —— 从空白开始容易卡在 layout 上
改哪个文件能改什么
references/shots/*.md镜头配方库的说明。Agent 就是读这些决定怎么拍assets/lib/PageCam.tsx镜头运动的实现。推拉摇移在这里assets/scripts/capture-template.mjs把产品网页截成贴图与 layout.json坑在哪一处
带 issue 编号的来自:github issues 按回复数排序 · 核验 2026-08-16
教程
没找到官方口径的教程 —— 官方只给了 README。
项目自述
以下是仓库 README 的开头部分(原文引用,抓取 2026-08-17)。读全文 ↗
video-shotcraft English | 中文 | 日本語 | Русский ▶ Browse all 150 shots in the Gallery · 🎬 Watch the Showcase What it does Tell Claude Code or Codex "make a promo for my product". The agent captures your real pages, picks shots from a library of tuned motion recipes, storyboards the film, animates it with Remotion, cuts it to the beat, adds sound design, checks every frame and hands you a finished MP4 you can keep editing. It is not a video generator. Every frame is drawn by code from your own screenshots and copy, so the result is sharp, on-brand and fully editable. How it works What's new > [!IMPORTANT] > ### 🎬 2026-10 · Shot library refresh > All 216 demos were polished and reviewed. The best versions now make up > 124 cards / 150 shots, with fresh Gallery previews and video-shotcraft branding. - 2026-09 · Motion Workbench. Keep editing the delivered film in a CapCut-style browser editor. - 2026-08 · video-talkcraft, the narration-video sibling: script + voiceover in, every motion beat locked to the voice (78 narration previews). - 2026-08 · JianYing (CapCut CN) export. The final film becomes an editable JianYing draft: per-shot clips, native caption tracks, separate SFX/BGM tracks. Verified on JianYing Pro 11.2 for macOS (guide). ❤️ Sponsors APIMart Thanks to APIMart for sponsoring this project! APIMart is a low-cost API platform for AI image & video generation — GPT-Image-2 from $0.006/image, 160+ images per dollar. One async API covers both image and video: submit a task, get an ID, fetch results via polling or callback. Batch tens of thousands of images without timeouts, switch models without changing code. Pay-as-you-go with no monthly fee — sign up here to get started. PackyCode Access leading AI models through PackyCode with one API endpoint and one API key. Enjoy fast, reliable access with automatic failover and dedicated high-speed routes for Codex and Claude Code. Get started with $1 in free credits, a discount on your first top-up, and savings of up to 80% on eligible routes. Pay in RMB with no currency conversion markups or extra top-up fees. Get started with PackyCode → AWS Amazon Web Services (AWS) offers reliable, scalable cloud computing services, including compute, storage and AI. Explore AWS → Install Built and tuned for Claude Code and Codex. The simplest way is to hand your agent the link: text Install this skill for me: https://github.com/Vincentwei1021/video-shotcraft Or with the skills CLI: bash npx skills add Vincentwei1021/video-shotcraft Or by hand: bash git clone https://github.com/Vincentwei1021/video-shotcraft.git cd video-shotcraft ln -s "$(pwd)" ~/.claude/skills/video-shotcraft # Claude Code ln -s "$(pwd)" ~/.codex/skills/video-shotcraft # Codex In mainland China, clone from the AtomGit mirror if GitHub is slow: git clone https://atomgit.com/VincentWei/video-shotcraft.git. Use Just describe the film. No command needed. | Say | What happens | | Use video-shotcraft to make a promo for my product. | The agent introduces the Ink Press template and asks whether to use it, then captures your pages and builds the film | | Make a launch film for this repo wi ……(后略,见原文)
判断
优势
- 画面确定、可复现 —— 代码渲染,不是抽卡;同样输入必然同样输出
- 镜头配方与可视化 Gallery 配套,可按画面选择。
- 自带完整可跑的样片模板(36.2 秒 / 10 镜头),不用从空白开始
- Apache-2.0;连它自己的 38 秒展示片都是用它做的
局限
- 只适合产品演示类:它的素材来自网页截图,不是任意实拍或生成画面
- 项目持续调整镜头库与工作台,使用时应以当前版本的配方和模板为准。
- 剪辑成本被用户提过(issue #37 在问有没有高性价比模型)
- 剪映草稿导出只在 mac 上测过,Windows 未测(README 自述)
不适用
要拍真实场景或人物的片子 —— 它的画面来自网页截图与代码渲染,不是这条路。
同管这些决定的其它项目
github README + GitHub API + DeepWiki 架构页 · 核验 2026-10-19 | 判断栏可报错。