Agent skill

Web Video Presentation

by ConardLi in ConardLi/garden-skills

Turns an article or spoken script into a click-through, full-screen 16:9 web presentation that looks like a video, with optional synthesized narration.

MITAuto-check passedMedia & Creative

SKILL.md written in Chinese; this summary is our English description.

Install Web Video Presentation

skills CLI
$ npx skills add ConardLi/garden-skills --skill web-video-presentation -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install ConardLi/garden-skills web-video-presentation --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/ConardLi/garden-skills.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/web-video-presentation .claude/skills/web-video-presentation && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
web-video-presentation
GitHub stars
13k
Token cost
~3.5k tokens
SKILL.md length
941 words
Files
130 (incl. scripts, references)
Skills in repo
4
Repo updated
First seen
Licence
MIT

At a glance

Turns an article or spoken script into a click-through, full-screen 16:9 web presentation that looks like a video, with optional synthesized narration.

  • Works in 4 steps: 内容编写(一次产出) → 网页开发 → 音频合成(可选) → …
  • Turning a spoken script or article into an explainer you can screen-record for a video platform
  • SKILL.md covers 适用场景, 工作流总览, 硬性自检协议(贯穿整个 Skill) and 各阶段文件读取指南, plus 9 more sections
  • Calls npm, bash and npx; needs OPENAI_API_KEY

What it does

The agent starts from your article or script and produces a narration script and an outline in one pass, then asks you to settle five things together: the script, the outline, a theme, your assets and how development should run (chapter by chapter, in sequence or in parallel). The outline fixes pacing and information density only; animation is designed per chapter while it is built.

The output is a Vite and React project in which each click advances one beat of the narration and every step fills the screen, with a progress bar that only shows on hover. A narrations.ts file is the single source for the step count and the audio, so the script, outline, chapter code, chapter list and audio files cannot drift apart.

Audio synthesis is optional and not tied to one vendor: MiniMax mmx-cli and OpenAI TTS are built in, and ElevenLabs, edge-tts, Azure or your own TTS can be swapped in. The script, outline and each finished chapter go through a checklist review, preferably by a separate reviewer agent, and failures are fixed before the agent reports back. The instructions are written in Chinese.

When your agent uses it

  • Turning a spoken script or article into an explainer you can screen-record for a video platform
  • Building a dynamic talk or keynote demo that has a cinematic feel and does not look like slides
  • Producing a narrated product walkthrough with one audio file per chapter
  • Recording a teaching tutorial as a 16:9 page with large type and one idea per screen

Example prompts

  • “Here is my voiceover script about our new API; build it as a click-driven web video I can screen-record.”
  • “Convert article.md into a 16:9 web presentation, and show me the outline before any code is written.”
  • “Generate the chapter audio for my web presentation with OpenAI TTS.”
  • “Develop the chapters of my-video in parallel now that I have approved the outline.”

Requirements

  • A source article or narration script
  • A text-to-speech provider such as MiniMax mmx-cli or OpenAI TTS, only if you want audio

Workflow steps

4 steps, taken from the step headings in SKILL.md.

  1. 内容编写(一次产出)
  2. 网页开发
  3. 音频合成(可选)
  4. 录屏 + 后期

What it can do on your machine

Read from SKILL.md and the folder at commit aaf9a82. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/, which the agent can run.

    Shell commands in SKILL.md call:

    • npm
    • bash
    • npx

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use npm and npx, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names these keys or tokens, usually read from environment variables:

    • OPENAI_API_KEY

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Web Video Presentation loads about 3.5k tokens when it runs, and up to ~32k if it reads all its reference files. Until then it costs about 135 tokens; SKILL.md has 941 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~135
When it runs · the whole SKILL.md, loaded when a task matches
~3.5k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~32k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from ConardLi/garden-skills at commit aaf9a82, republished under its MIT licence (© ConardLi). 941 words, ~3,509 tokens.

Download SKILL.mdSave it as .claude/skills/web-video-presentation/SKILL.md (or your agent's skills folder). This skill also uses 129 other files; get the full folder from GitHub.
name
web-video-presentation
description
把一篇文章或口播稿,做成"看起来像视频"的点击驱动 16:9 网页演示,可选合成口播音频。流程:原始文章 → **一次产出**口播稿 + outline 开发计划 → 用户**一次对齐** 5 件事(稿子 / outline / 主题 / 素材 / 开发模式)→ 网页开发(逐章 / 顺序 / 并行)→ 可选音频合成(provider-agnostic:内置 MiniMax mmx-cli + OpenAI TTS,可换 ElevenLabs / edge-tts / Azure / 自带 TTS)。**outline 只规划节奏与信息密度,不规划动画** —— 动画由章节开发时按 PRINCIPLES + ANTI-AI 法则即时设计。每次点击推进口播稿的一个节拍,每一步独占整屏,进度条平时隐藏只在悬浮时出现。适用场景:用网页做视频(动态 PPT 但不像 PPT)、把口播稿 / 文章变成可交互的解说、为 B 站 / YouTube / 视频号录屏教程、做有电影感的产品 / talk demo。本 Skill 沉淀的是设计方法论 + 协作流程 —— 不绑定任何特定样式 / 字体 / 颜色 —— 因此能复用到任意主题与美学。

Web Video Presentation

把一篇文章或口播稿,一步步做成可录屏的"伪装成视频的网页",可选合成 口播音频。产出物 = Vite + React + TS 项目 + 按章节切分的音频。

适用场景

  • "我有口播稿 / 一篇文章,帮我做成视频" —— 口播驱动的内容
  • 想做 "动态 PPT"
  • 16:9 横屏录屏,大字、留白、每屏都要有动效
  • 教学 / 产品演示 / keynote 想要电影感
  • B 站 / YouTube /抖音视频内容

本 Skill 以方法论 + 协作流程为核心。脚手架模板提供 token 和原语, 但每个美学决策(配色、字型、动效气质)都应该针对你的主题重新设计 —— 不要照搬。


工作流总览

Phase 1   内容编写
   1.1  识别用户输入
   1.2  一次产出 script.md + outline.md
        (口播稿 + 开发计划)
   ▼
[Checkpoint Plan]      ← 必须停。一次对齐 5 件事:
                         稿子 / outline / 主题 / 素材 / 开发模式
   ▼
Phase 2   网页开发
   2.1  脚手架(按选定主题)
   2.2  第 1 章 = 主线程 + 完整版本(强制 anchor)
        ▼
        [硬节点] 用户验收第 1 章 ← 不可跳过
        ▼
   2.3  第 2~N 章(按选定模式:A 逐章 / B 顺序 / C 并行)
   ▼
[Checkpoint Audio]     ← 必须停。是否合成音频
   ▼
Phase 3   音频合成(可选)
   ▼
Phase 4   录屏 + 后期

工作目录约定(agent 在用户当前目录下创建 / 编辑):

my-video/
├── article.md          # 用户给原文时必有 —— 不删!开发阶段画面信息源
├── script.md           # 必有:保持原文语言的平台化口播稿(决定节拍)
├── outline.md          # 必有:开发计划(章节切分 + 每步内容 + 信息池)
└── presentation/       # 脚手架产出的 Vite + React + TS 项目
    ├── src/chapters/<NN>-<id>/
    │   ├── <Chapter>.tsx     # 视觉实现
    │   ├── <Chapter>.css
    │   └── narrations.ts     # ★ step 数 + 口播文本的唯一真相源
    ├── scripts/
    │   ├── extract-narrations.ts   # 扫所有 narrations.ts → audio-segments.json
    │   ├── synthesize-audio.sh     # provider-agnostic runner(循环 segments)
    │   └── tts-providers/          # 每 provider 一个 .sh(内置 2 个)
    │       ├── README.md           # 三函数契约 + 5 段现成代码片段(11labs / edge-tts / say / azure / gcloud)
    │       ├── minimax.sh          # 默认 provider,用 mmx-cli
    │       └── openai.sh           # 内置 OpenAI TTS(curl + OPENAI_API_KEY)
    ├── audio-segments.json         # extract 产出(合成前 review)
    └── public/audio/<id>/<N>.mp3   # 可选:合成的音频

关键:narrations.ts 是 step 数和音频合成的唯一真相源。 章节 .tsx 里的 if (step === N) 出现的最大 N + 1 必须等于 narrations.length。这保证 5 处地方(script / outline / 章节代码 / chapters.ts / 音频文件)永远不会漂。


硬性自检协议(贯穿整个 Skill)

下面三个产出,每一个完成后必须走自检 → 修复 → 再汇报 / 推进:

产出自检清单出处
script.mdSCRIPT-STYLE.md 三层自检(形式 / 风骨 / 念出来)
outline.mdOUTLINE-FORMAT.md 自检
单章实现完成CHAPTER-CRAFT.md 完工自检

执行方式(按能力降级,优先用更隔离的方式):

  1. Agent Teams(最优):开一个独立的 reviewer agent,给它"产出文件 路径 + 对应清单 + 关键上下文",让它逐项核查并严格汇报结论 (哪几条 pass / 哪几条 fail + 证据 + 改写建议)。
  2. subAgent(次优):没有 Teams 能力但能开 subagent 就用 subagent 走同样流程。
  3. 自检(兜底):当前 agent 都没有上述能力,就自己严格逐项 核查 —— 不允许目测一遍就放行。

铁律:拿到结论后先按 fail 项把产出改完,再向用户汇报"做完了

  • 自检结论 + 改了什么"。直接拿原始结论汇报但不修复 = 违规。

各阶段文件读取指南

不同阶段读不同的文件。长会话里 agent 容易遗忘原则,特别是 Phase 2.4 的"实现单章"会重复 N 次 —— 每次都要回看核心约束。

阶段必读(每次都看)一次性看完 / 按需查
Phase 1.1-1.2 内容编写references/SCRIPT-STYLE.md + references/OUTLINE-FORMAT.md + article.md(用户原文,如有)——
Checkpoint Plan 选主题——themes/*/theme.json(动态读全部,列清单 + bestFor 推荐 + descriptionZh);references/THEMES.md(用户想了解主题系统时)
Phase 2.1 脚手架——SKILL.md 本节看一次
Phase 2.4 实现单章(×N 次,被 2.2 / 2.3 调用)references/CHAPTER-CRAFT.md 单一入口 —— Part 0 十条原则 / Part 1 开工 5 问 / Part 2 关系→动作决策树 / Part 3 视觉工具箱 / Part 4 时长参考 / Part 5 反 AI 味反模式 / Part 6 代码硬规则(含 narrations.ts 强制约束)/ Part 7 完工自检 / Part 8 反馈速查 + 当前主题的 themes/<id>/theme.json + 当前章节的 outline.md 段落 + article.md 本章对应段落 + 素材清单references/EXAMPLES/(结构示意,不是抄袭模板);references/THEMES.md 完整 token 契约
Phase 3 音频合成references/AUDIO.md(含 narrations.ts → segments.json → 任意 provider 流程,内置 minimax + openai)templates/scripts/tts-providers/README.md(换 provider / 自带 TTS 时)
Phase 4 录屏 + 后期references/RECORDING.md(含 ?auto=1 自动录屏)——
选 / 造 / 切主题——references/THEMES.md

写章节时只读一份 CHAPTER-CRAFT.md。十条原则 / 开工 self-prompting / 决策树 / 反 AI 味反模式 / 完工自检全部并入这一份单一入口。EXAMPLES/ 不是必读 —— 先按内容自由设计,卡壳才翻(按 anchor 翻"形",不要照搬)。


Phase 1 —— 内容编写(一次产出)

1.1 识别用户输入
用户给的东西该做的
原始文章(书面语 / 公众号 / 论文 / 博客)一次产出 script.md + outline.md(1.2),过 Checkpoint Plan
直接的口播稿 / 视频脚本落盘成 script.md,一次产出 outline.md(1.2 简化版),过 Checkpoint Plan
啥都没有,只说"帮我做个 X 主题的视频"反问:先给一段素材或大纲。Skill 不替用户构思内容
1.2 一次产出 script.md + outline.md

两份产出物在一次思考中完成:

  1. 生成 script.md:按 references/SCRIPT-STYLE.md 的规则把 article 转成保持原文语言的平台化口播稿。保留 article.md 不删——它是 outline 写信息池和章节实现画面时的细节源(双源原则)。
  2. 生成 outline.md:按 references/OUTLINE-FORMAT.md 规则切章节 + 切 step + 每章首段抽信息池。

outline 的边界(关键):

outline 必须写outline 不要写
章节切分 / 每章 step 数 / 估时具体动画类型(blur clear / wipe / 弹簧)
每步屏幕内容(hero / 数据 / 标语 / 列表项)CSS 实现手段(filter / SVG / clip-path)
章节级信息池:从 article 抽的数字 / 引用 / 案例 / 标签时长数值(不写 2.5s / 80120ms)
步级关系名前缀("反差对照" / "递进列表" / "金句" 等可选 hint)持续微动 / 错峰量等微观节奏

outline 不写动画的理由:写死动画 = chapter agent 退化为翻译机; 留白让 chapter agent 在每步开工时按 CHAPTER-CRAFT.md 的"内容驱动决策树"自由设计,才有真正的视频感。详见 CHAPTER-CRAFT.md Part 0 原则 7。

落盘后必须先走自检再进 Checkpoint Plan:按上文「硬性自检协议」分别 对 script.md / outline.md 执行(优先 Agent Teams → subAgent → 自检), 按结论修复完成后再进入 Checkpoint Plan。


Checkpoint Plan —— 5 件事一次对齐(硬节点)

script.md + outline.md 写完后必须停下来。用户在这一个节点同时确认 5 件事。

agent 此时要做的预备工作
  1. 读所有 themes/*/theme.json 拿 nameZh / descriptionZh / bestFor / mood —— 不要硬编码清单
  2. 根据 script.md 的内容类型 / 关键词 / 语气,主动从主题里挑 2~3 套最匹配的推荐(匹配 bestFor 字段)
  3. 扫一遍 outline.md 末尾"素材清单"部分
总结模板(骨架,agent 按情况填充)
内容计划写完,产出文件:
  📄 article.md     {若用户给原文则保留}
  📄 script.md      {X} 字 / ~{T} 分钟
  📄 outline.md     {N} 章 / {M} 步 + 每章信息池 + 末尾素材清单

章节速览:
  1. <id>     <章节标题>    <S> 步 ~<T>s
  2. ...

接下来一次对齐 5 件事:

  1. 稿子 (script.md) 要不要改?
     可以直接编辑文件,或口头告诉我修改方向。

  2. 开发计划 (outline.md) 要不要改?重点看:
     - 章节切分 / step 数 / 估时是否合理(合理判断:每章 30~60s)
     - 每步屏幕内容是否清晰
     - 每章首段「信息池」是否有足够的 article 细节供画面挂
     - 末尾素材清单是否完整

  3. 选哪个主题?我的推荐:
     ★ <推荐 1:nameZh (id)> — 因为 <bestFor 命中>;<descriptionZh 摘要>
     ★ <推荐 2 / 推荐 3>
     其它可选:<剩余主题,nameZh + 一句话>
     也可以让我帮你做新主题(详见 references/THEMES.md)。

  4. 真素材怎么准备?粗看本视频要的图:<列粗略清单>
     a) 我从 <现有素材路径> 帮你挑   b) 你自己提供   c) 全部 placeholder

  5. 开发模式选哪个?

     **第 1 章无论哪种模式都必须主线程做完 + 用户验收**(强制 anchor)。
     差异在第 2 章及之后:

     A) 默认 · 逐章确认(推荐)
        每章做完都暂停验收 → 风险可控 / 节奏最稳
     B) 第 1 章后顺序开发(不并行)
        第 2~N 章主线程顺序做完后统一验收 → 速度中 / 适合 agent 不支持并行
     C) 第 1 章后并行开发(subagent)
        第 2~N 章用 subagent 并行 → 最快 / 用户控并行数(一次几章)
        ⚠️ 风格各章会有差异(这是预期,主题禁区兜底)

收到反馈后:

  • 稿子 / outline 要改:直接编辑文件,编辑完 ping 一次(或口头描述 agent 改)
  • 主题必须明确才进入 Phase 2。用户说"主题你帮我选" → 取你推荐的第 1 个, 告诉用户你选了什么、为什么,给反悔机会
  • 模式选定 → 进 Phase 2

Phase 2 —— 网页开发

2.1 脚手架
bash
bash <path-to-web-video-presentation>/scripts/scaffold.sh \
  ./presentation \
  --theme=<用户选的主题 id>

bash <path-to-web-video-presentation>/scripts/scaffold.sh --list-themes

自定义主题 → 先按 references/THEMES.md "创作新主题"流程做一个 themes/<my-theme>/,再 --theme=<my-theme>。

脚手架带一个 01-example demo。在写第一章真实内容前删掉:

bash
rm -rf presentation/src/chapters/01-example

并把 presentation/src/registry/chapters.ts 里 EXAMPLE_CHAPTER 的 import 和数组项移除。

2.2 第 1 章 —— 主线程 + 强制验收

核心:第 1 章 = 完整版本一次到位(节奏 + 视觉 + 真素材齐全)。 没有"骨架版"概念 —— 第一章就要做出用户能直接验收的样板。

为什么第 1 章必须主线程:

  • 它是 CHAPTER-CRAFT.md 这套指引在当前 主题 + 当前题材下的第一次落地
  • 如果指引有盲区 / 主题颜色 / 字体 token 不够用,第 1 章一定会暴露 —— 这时候有人类反馈就能修指引 / 调主题,早改成本最低
  • 后续章节(无论顺序 / 并行)都要参考第 1 章的代码模式,所以第 1 章 = 当次项目的"风格锚点(不强求章节间一致,但单章自身得有完整说服力)"

做完第 1 章后必须停下来等用户验收:

第 1 章 <id> 做完了,dev server 在 localhost:5173 运行。

验收重点:
  □ 视觉气质对不对?符合 <theme nameZh> 的预期吗?
  □ 节奏对不对?某些步太快 / 太慢 / 信息太薄?
  □ 内容驱动动画是否到位?还是有几步是无脑入场动画?
  □ 双源原则:屏幕画面有没有"口播没念但 article 能挂"的细节?
  □ 反 AI 味检查:紫粉渐变 / 圆角彩色边框 / 假插画 / emoji 是否有?

问题告诉我,我针对性改。OK 了告诉我"继续",我按选定模式做第 2 章及之后。
2.3 第 2~N 章 —— 按选定模式

所有模式下的共同规则:每章独立按 CHAPTER-CRAFT.md 开发。风格不强求章节间完全一致 —— 主题颜色 / 字体 token 兜底视觉 统一,动画 / 节奏 / 视觉演示由章节自由发挥是设计预期。

模式 A · 默认 · 逐章确认

第 2 章做完 → 暂停验收 → OK → 第 3 章 → 暂停 → ... → 第 N 章。每章 独立验收,问题随时改,风险最低,节奏最稳。用户不明确选模式时 默认走这个。

模式 B · 第 1 章后顺序开发

第 2 章 → 第 3 章 → ... → 第 N 章 主线程顺序做完,最后统一验收。 速度中等,适合 agent 不支持并行任务的环境。

Show full SKILL.md (411 more words)Show less
模式 C · 第 1 章后并行开发(subagent)

用 subagent 把第 2~N 章并行做完,最大并行数由用户控制("一次 4 章" / "一次 2 章")。最快,但风格各章会有差异 —— 这是预期,因为:

  1. 每个 subagent 看不到别的 subagent 产出,无法机械对齐
  2. 章节代码物理分离(每章一个文件夹 / 自己的 CSS 前缀),不会互相 破坏
  3. 主题 token 兜底视觉统一(颜色 / 字体 / hero 数字 / 卡片 / 分割线 性格 / 装饰),气质不会跑偏
  4. 风格不一致 = 人手写视频的呼吸感(多 voice / 多视角)

并行 subagent 的 prompt 必须包含:

  • 当前章节 outline 段落(含信息池)
  • references/CHAPTER-CRAFT.md 的路径(单一必读 —— 视觉演示要求 + 逐步揭示 + 双源原则 + 反 AI 味 + 代码红线 + 完工自检全部在这一份里)
  • 当前主题 theme.json 的 descriptionZh / mood / bestFor(参考气质 即可,动画 / 时长 / 字号 / emoji 由 chapter agent 自由决定)
  • 第 1 章代码作为"代码风格"参考(不是"视觉抄袭对象")
  • 硬规则:每章独立 CSS 前缀(.cd- / .mg- / .pm- / ...); 不修改 chapters.ts;完工跑 npx tsc --noEmit

重要:无论选哪种模式,用户随时可以中途切换模式。第 2 章 OK 后用户说"剩下的并行" / "剩下的逐章" 都行。

2.4 实现单章(每章必走)

详细指引见 references/CHAPTER-CRAFT.md —— 单一必读入口,覆盖:视觉演示要求 / 逐步揭示 / 内容取舍 / 双源原则 / 视频演示基本审美 / 反 AI 味 / 代码红线 / 完工自检。

核心要点(CHAPTER-CRAFT.md 详述):

  • 每章必须有 CSS / SVG / Canvas / JS 视觉演示,禁纯文字章节
  • 逐步揭示:清单 / 列表必须 1 项 = 1 step,禁一次全展示
  • 双源原则:节奏跟口播稿(顺序不能乱),细节回原文章抽(信息池 + 本章 article 段落)
  • 完工自检逐项过,不达标回去改 —— 按上文「硬性自检协议」执行 (优先 Agent Teams → subAgent → 自检),改完再向用户汇报本章交付
2.5 大改后 bump STORAGE_KEY

改动 chapters.ts(增加 / 删除 / 重排章节,或某章 narrations.ts 长度变化)后,bump presentation/src/hooks/useStepper.ts 的 STORAGE_KEY(如 v4 → v5),避免持久化游标落到不存在的 step 上。


Checkpoint Audio —— 是否合成音频(硬节点)

Phase 2 结束后必须停下来,问用户:

网页做完,{N} 章 {M} 步,dev server 在 localhost:5173 跑着。

要不要合成音频做"自动播放录屏"?
  ✓ 合成 → 扫所有章节的 narrations.ts 出 audio-segments.json,
           调 TTS provider 合成每步一个 mp3 到 public/audio/。
           合成完后用 ?auto=1 模式可以一镜到底录屏(音视频天然同步)。
           内置两个 provider:
             • minimax (mmx-cli)    —— 默认,中文音色稳
             • openai  (OPENAI_API_KEY) —— curl-based,多数已有 key
           其它后端 (ElevenLabs / edge-tts 免费 / macOS say 离线 /
           Azure / Google) 见 scripts/tts-providers/README.md 的现成片段。
  ✗ 不合成 → 跳过 Phase 3,直接 Phase 4 用手动录屏 + 后期配音。

要合成 → Phase 3。不合成 → 直接 Phase 4。


Phase 3 —— 音频合成(可选)

详细流程见 references/AUDIO.md。简版:

bash
cd presentation
npm run extract-narrations   # 扫所有 narrations.ts → audio-segments.json
# 让用户扫一眼 audio-segments.json 确认文本对
npm run synthesize-audio                       # 默认 minimax provider,增量
# 或用内置 openai (要 OPENAI_API_KEY):
PRESENTATION_TTS=openai npm run synthesize-audio
# 或自定义:写一个 scripts/tts-providers/<name>.sh,见该目录的 README.md

合成完告诉用户:输出位置 / 总段数 / 哪些段时长异常(太长 = 该 step 拆 分;太短 = 文案太薄)—— 给最后一次校准节奏的机会。然后进入 Phase 4。


Phase 4 —— 录屏 + 后期

详见 references/RECORDING.md。两种路径:

场景推荐路径
Phase 3 已合成音频Auto 模式一镜到底:浏览器开 localhost:5173/?auto=1 → 按 SPACE → 整片自动播完 → 停录 → 裁头尾即成片,无需后期对音轨
Phase 3 跳过默认 Manual 模式手动点击推进 → 后期任意剪辑工具配音

agent 在 Phase 3 / Checkpoint Audio 后主动告诉用户适合的录屏路径。


十条原则(一句话清单)

完整展开见 references/CHAPTER-CRAFT.md Part 0 —— 写章节时回那里查,下面只是索引。

#原则一句话
116:9 固定舞台内容 1920×1080 + transform scale,没有响应式
2全局 step 计数器章节是 step 的纯函数,无定时器
3每步独占整屏if (step === N) return <FullScene />
4口播节拍 = step一节拍 = 一 step = 一聚焦想法
5隐藏的边角控件进度条 / 翻页器默认 opacity 0
6舞台无 chrome没有 header / footer / 页码 / 品牌条
7内容驱动动画先找内在动作,找不到才入场动画兜底;持续微动慎用
8多点逐个揭示1 项 = 1 step,禁同步 stagger 上 N 项
9整片同一主题章节间不翻表面色;颜色 / 字体走 token,其它尺度章节自由
10双源原则script 定节拍,article 定画面密度(落到信息池)

常见用户反馈速查

简化表见 references/CHAPTER-CRAFT.md Part 8「常见反馈速查」。关键:先定位是哪一层(节奏 / 视觉 / 内容 / 代码),再改最小切片,不要重做整章。


相关资源

按"何时读"标注,避免一次性全读:

文件何时读内容
references/SCRIPT-STYLE.mdPhase 1.2 必读文章 → 口播稿规则、平台变体
references/OUTLINE-FORMAT.mdPhase 1.2 必读outline.md 字段 spec、命名约定、章节切分、信息池
references/CHAPTER-CRAFT.mdPhase 2.4 每章单一必读入口Part 0 十条原则 / Part 1 开工 5 问 / Part 2 关系→动作决策树 / Part 3 视觉工具箱 / Part 4 时长 / Part 5 反 AI 味反模式 / Part 6 代码硬规则 / Part 7 完工自检 / Part 8 反馈速查
references/EXAMPLES/可选 —— 看结构章节结构示意(hook / list-reveal / case-tech-review);不是抄袭模板
references/THEMES.md选 / 造 / 切主题时完整 token 契约 + 内置主题清单 + 创作流程
references/AUDIO.mdPhase 3 才读provider-agnostic 音频合成流程、内置 minimax 用法、换 provider 路径、故障排查
templates/scripts/tts-providers/README.md换 / 加 TTS provider 时三函数契约 + 内置 2 个 (minimax / openai) + 5 种现成代码片段(ElevenLabs / edge-tts / macOS say / Azure / Google)
references/RECORDING.mdPhase 4 才读录屏工具 + 后期合成
themes/Checkpoint Plan / Phase 1.2 时翻内置主题(每个含 theme.json + tokens.css)
scripts/scaffold.shPhase 2.1 跑一次一键项目脚手架

© ConardLi, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 129 other files (scripts, references) in skills/web-video-presentation of ConardLi/garden-skills.

  • SKILL.md
  • README.md
  • README.zh-CN.md
  • manifest.json
  • references/AUDIO.md
  • references/CHAPTER-CRAFT.md
  • references/EXAMPLES/README.md
  • references/EXAMPLES/case-tech-review/README.md
  • references/EXAMPLES/case-tech-review/outline-snippet.md
  • references/EXAMPLES/hook-chapter/README.md
  • references/EXAMPLES/hook-chapter/chapter.css
  • references/EXAMPLES/hook-chapter/chapter.tsx
  • references/EXAMPLES/list-reveal/README.md
  • references/EXAMPLES/list-reveal/chapter.css
  • references/EXAMPLES/list-reveal/chapter.tsx
  • references/OUTLINE-FORMAT.md
  • … and 114 more

Open the folder on GitHubat commit aaf9a82

Compare with similar skills

Web Video Presentation next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Web Video Presentation compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Web Video Presentation this skillConardLi/garden-skills13k—~3.5kAutomated safety check: PassMIT
Voice Clone Ttsnpc-live/clawfirm156—~1.1kAutomated safety check: PassNone
KrillinAI CLI Operatorkrillinai/OpenCreator13k—~869Automated safety check: PassApache-2.0
Video Productionspeechlab0210/video-production-skill105—~4.1kAutomated safety check: NotesMIT
Super Video MakerBomx/super-video-maker-skill308—~11kAutomated safety check: NotesNone
Motion Videobestagentkits/motion-video-skill113—~1.5kAutomated safety check: PassMIT

Similar skills

  • Voice Clone Tts

    npc-live/clawfirm

    声纹克隆和语音合成。上传音频样本克隆声纹,用克隆声纹或预设声纹生成语音。支持多个后端:MiniMax、ElevenLabs、Fish Audio、Azure TTS、OpenAI TTS。支持情绪控制、语速调整、批量生成。触发词:语音合成、TTS、声纹克隆、voice clone、text to speech、配音、旁白。

    156 GitHub stars~1.1k tokensUpdated 3 mo ago
    Media & CreativeAuto-check passed
  • KrillinAI CLI Operator

    krillinai/OpenCreator

    Routes agents to the right KrillinAI command for subtitles, dubbing, video rendering, covers and speech, and explains how to read its JSON and manifest output.

    13k GitHub stars~869 tokensUpdated 3 days ago
    Media & CreativeAuto-check passed
  • Video Production

    speechlab0210/video-production-skill

    AI educational video production pipeline. An agent skill from speechlab0210/video-production-skill.

    105 GitHub stars~4.1k tokensUpdated 3 mo ago
    Media & CreativeAuto-check: notes
  • Super Video Maker

    Bomx/super-video-maker-skill

    End-to-end AI video production skill for agentic frameworks.

    308 GitHub stars~11k tokensUpdated 2 mo ago
    Media & CreativeAuto-check: notes
  • Motion Video

    bestagentkits/motion-video-skill

    Produce beat-synced 1080p motion-graphic videos in HyperFrames (HTML + GSAP) with an AI voice-over, Vietnamese karaoke captions, SFX and generated music, in one of two proven styles (glass keynote…

    113 GitHub stars~1.5k tokensUpdated 10 days ago
    Media & CreativeAuto-check passed
  • Render Hand Swipe Flavor Listicle

    gooseworks-ai/goose-skills

    Render a 'hand-swipe flavor listicle' video from a config — the brand's REAL product cutouts each float on a flat flavor-matched color field under a persistent title and slide RIGHT-TO-LEFT one to…

    1.2k GitHub stars~2.9k tokensUpdated today
    Media & CreativeAuto-check passed

More from ConardLi/garden-skills

  • Beautiful Article

    ConardLi/garden-skills

    Turns a URL, PDF, DOCX, Markdown file, text or screenshots into a designed, shareable single-file HTML article through a staged review workflow.

    13k GitHub stars~4.7k tokensUpdated 2 mo ago
    Auto-check passed
  • GPT Image 2 Prompt and Generation

    ConardLi/garden-skills

    Generates or edits images with GPT Image 2 through an OpenAI-compatible API, or falls back to writing finished prompts from a library of structured templates.

    13k GitHub stars~4k tokensUpdated 2 mo ago
    Auto-check: notes
  • Web Design Engineer

    ConardLi/garden-skills

    Builds polished browser-rendered HTML, CSS, JavaScript and React pages, dashboards, prototypes and slide decks, aiming for a stunning rather than merely functional result.

    13k GitHub stars~8.8k tokensUpdated 2 mo ago
    Auto-check passed

Questions about Web Video Presentation

What does Web Video Presentation do?

Turns an article or spoken script into a click-through, full-screen 16:9 web presentation that looks like a video, with optional synthesized narration. The agent starts from your article or script and produces a narration script and an outline in one pass, then asks you to settle five things together: the script, the outline, a theme, your assets and how development should run (chapter by chapter, in sequence or in parallel). The outline fixes pacing and information density only; animation is designed per chapter while it is built.

When should I use Web Video Presentation?

Web Video Presentation fits situations like: turning a spoken script or article into an explainer you can screen-record for a video platform; building a dynamic talk or keynote demo that has a cinematic feel and does not look like slides; producing a narrated product walkthrough with one audio file per chapter; recording a teaching tutorial as a 16:9 page with large type and one idea per screen.

How do I install Web Video Presentation in Claude Code?

Run `npx skills add ConardLi/garden-skills --skill web-video-presentation -a claude-code`. Or copy the skill folder (skills/web-video-presentation in ConardLi/garden-skills) into .claude/skills/web-video-presentation in your project. Claude Code loads it when a task matches its description.

How do I install Web Video Presentation in Codex?

Run `npx skills add ConardLi/garden-skills --skill web-video-presentation -a codex`. Or copy the skill folder (skills/web-video-presentation in ConardLi/garden-skills) into .agents/skills/web-video-presentation in your project. Codex loads it when a task matches its description.

Can I use Web Video Presentation in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add ConardLi/garden-skills --skill web-video-presentation -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/web-video-presentation, .gemini/skills/web-video-presentation, .github/skills/web-video-presentation and .opencode/skills/web-video-presentation in your project.

What does Web Video Presentation need to run?

Going by SKILL.md and its folder, Web Video Presentation needs the command-line tools its instructions call (npm, bash and npx) and credentials named OPENAI_API_KEY. Our summary lists: A source article or narration script; A text-to-speech provider such as MiniMax mmx-cli or OpenAI TTS, only if you want audio.

Does Web Video Presentation access the network?

SKILL.md contains no URLs. Its commands use npm and npx, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Web Video Presentation safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Web Video Presentation use?

Web Video Presentation is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Web Video Presentation use?

About 3.5k tokens (SKILL.md is roughly 14k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 28k tokens, read only when the agent opens those files.

What are the alternatives to Web Video Presentation?

Skills that share tags, products or a category with Web Video Presentation: Voice Clone Tts (npc-live/clawfirm, 156 stars), KrillinAI CLI Operator (krillinai/OpenCreator, 13k stars), Video Production (speechlab0210/video-production-skill, 105 stars) and Super Video Maker (Bomx/super-video-maker-skill, 308 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Web Video Presentation?

ConardLi (a GitHub user) maintains it in ConardLi/garden-skills, which has 12,767 GitHub stars. The repository holds 4 skills in this directory. The repository was last updated on July 12, 2026.

Source: ConardLi/garden-skills on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.