Media Toolkit

AI-powered short video production pipeline — from idea to published video, driven by Claude Code skills.

中文: AI 驱动的短视频生产流水线 — 从创意到发布，全部由 Claude Code 技能驱动。

Designed for Douyin (TikTok China) vertical format (1080x1920, 30fps). Uses Remotion for programmatic video generation.

中文: 面向抖音竖屏格式（1080x1920, 30fps），使用 Remotion 进行程序化视频生成。

English | 中文

Showcase / 作品展示

影视飓风 AI 工作流 (2026-04-28)

拆解 1600 万粉影视飓风的 AI 工具链 — 从调研到成片，全程 AI 驱动。

封面：

配音： voiceover-full.mp3 (6 镜头完整配音)

渲染视频： ying-shi-ju-feng-ai-workflow.mp4 (1080x1920 竖屏 30fps, 5.4MB)

查看脚本与代码

分镜脚本 — 6 镜头，69 秒
口播文案
调研文档
审核报告
Remotion 组合 — 6 个镜头组件

DeepSeek V4 (2026-04-25)

成品	文件	说明
渲染视频	`projects/2026-04-25-deepseek-v4/output/deepseek-v4.mp4`	1080×1920 竖屏 30fps

中文

流水线概览

创意 → 脚本 → 审核 → 素材 → 代码 → 审核 → 渲染 → 发布
 │      │      │      │      │      │      │      │
 │  /video- /script- TTS  /remotion- /video- render /douyin-
 │  script  review  封面  video     review       publish

步骤	命令	说明
1. 脚本	`/video-script <创意>`	联网调研，生成分镜脚本（画面、口播、字幕）
2. 审核	`/script-review <slug>`	逐条事实核查，质量评估，自动修复
3. 配音	`/voiceover <slug>`	火山引擎 TTS 生成配音，逐镜头 MP3 + Whisper 对齐字幕时间轴
4. 封面	`/video-cover <slug>`	AI 生成背景图 + Python 叠加文字，输出 4:3 横版 + 3:4 竖版封面
5. 代码	`/remotion-video <slug>`	将分镜脚本转为 Remotion React 组合（布局原语、动画、转场、音频）
6. 审核	`/video-review <slug>`	审计 Remotion 代码：抖音合规、动画质量、脚本一致性，自动修复
7. 渲染	`npx remotion render`	导出 MP4 视频
8. 发布	`/douyin-publish <slug>`	从脚本自动提取标题和标签，发布到抖音

快速开始

环境要求：

Claude Code CLI（技能在 Claude Code 中运行）
Node.js 18+ 和 pnpm
Python 3.7+（用于封面文字叠加）

安装：

git clone https://github.com/yasinshaw/media.git
cd media
cp .env.example .env  # 填入下方 API 密钥
cd remotion && pnpm install

API 密钥配置：

复制 .env.example 为 .env，填入以下密钥：

密钥	用途	获取方式
`IMAGE_API_KEY` + `IMAGE_API_BASE_URL` + `IMAGE_MODEL`	AI 图片生成（OpenAI 兼容接口，可配置任意供应商）	中转站 / OpenAI / Gemini 中转，详见 `.env.example`
`VOLC_TTS_API_KEY`	TTS 配音合成	火山引擎 TTS 控制台 → API 密钥管理
`TAVILY_API_KEY`	联网调研搜索	Tavily 注册 → API Keys 页面

只需配置你用到的功能对应的密钥。例如只生成脚本不需要配音和封面，只需 TAVILY_API_KEY。

制作第一个视频：

# 1. 从创意生成分镜脚本
/video-script DeepSeek V4 发布了，推理能力怎么样

# 2. 事实核查
/script-review deepseek-v4

# 3. 生成配音
/voiceover deepseek-v4

# 4. 生成封面
/video-cover deepseek-v4

# 5. 转为 Remotion 代码（Studio 预览）
/remotion-video deepseek-v4

# 6. 审核代码 + 渲染 MP4
/video-review deepseek-v4 --render

# 7. 发布到抖音
/douyin-publish deepseek-v4

技能详情

`/video-script` — 分镜脚本生成

将视频创意转化为完整的分镜脚本：

调研阶段: 搜索官方文档、GitHub 仓库、权威媒体
角度选择: 提出 3 个创意角度（创意明确时跳过）
脚本生成: 钩子 → 痛点 → 核心内容 → CTA
时长计算: 中文语速 4-5 字/秒，根据字数反推时长
画面类型: remotion、实景拍摄、固定图片、ai生图、ai生视频、ai背景图
音频标注: BGM 风格/节奏，每个镜头的音效

输出: script.md、voiceover.md、research.md

`/script-review` — 事实核查与质量审核

严格的审核流程，附来源验证：

逐条核查事实性声明（统计数据、产品信息、日期）
验证时长（字数 vs. 镜头时长）
自动修复：未经验证的数据、不正式用语、夸张表述
需确认后执行：重大事实修正、叙事调整

`/voiceover` — TTS 配音生成

通过火山引擎 TTS API 生成配音：

默认音色：刘飞 2.0（成熟男声，适合科技内容）
逐镜头分文件输出，便于后期剪辑
Whisper 对齐的字幕时间轴清单
完整合并文件用于快速预览

`/video-cover` — 封面图生成

AI 生成多平台封面：

4:3 横版（B站、西瓜视频）
3:4 竖版（抖音、快手、视频号）
AI 生成纯背景 → Python 叠加文字，支持 AI 推荐配色

`/remotion-video` — Remotion 代码生成

将分镜脚本转为生产级 React 视频代码：

布局原语: CenteredStack、HubLayout、TwoColumnCompare、TimelineFlow
动画: Spring 预设、interpolate、打字机效果、关键词高亮
转场: TransitionSeries（淡入淡出、滑动、擦除、翻转、时钟擦除）
音频: 逐镜头配音、背景音乐、音效
AI 背景: 火山引擎 API 生成电影感镜头背景
视觉效果: 光线泄露、动态模糊、星芒放射、噪点纹理、Lottie、动态图表

输出: 镜头组件 + 组合 + Remotion Studio 预览

`/video-review` — 代码审核与渲染

Remotion 组合的自动化质量门禁：

Remotion 最佳实践合规（禁用 CSS 过渡、音频必须用 staticFile() 等）
抖音格式合规（1080x1920、安全区域、字幕定位）
脚本与代码一致性（镜头数、时长、字幕文本）
视觉对齐（SVG 坐标、absolute/flex 混用、z-index）
动画流畅度（禁止布尔值 opacity 跳变）
自动修复大部分问题；支持 --render 一键审核+渲染

`/douyin-publish` — 抖音发布

自动发布到抖音：

从项目文件提取标题、标签和封面
根据内容类型生成话题标签
通过 auto-douyin API 发布

Remotion 组件

组件	用途
`CenteredStack`	垂直内容堆叠，内置安全区域 + 字幕
`HubLayout`	中心节点 + 8 个周围位置，自动 SVG 连线
`TwoColumnCompare`	左右对比（纵向/横向）
`TimelineFlow`	顺序流程，带徽章和连接器
`SafeArea`	抖音安全区域包装器（上 120px、下 200px、左右 40px）
`ProgressiveSubtitle`	逐词字幕，Whisper 对齐时间轴
`BGMAudio`	背景音乐，风格映射 + 音量控制
`SFXLayer`	音效层（whoosh、impact、text-pop、outro）

外部服务

服务	用途	环境变量
OpenAI 兼容图片生成	图片生成（中转站 / OpenAI / Gemini 中转，可配置）	`IMAGE_API_KEY` / `IMAGE_API_BASE_URL` / `IMAGE_MODEL`
火山 TTS	配音合成（TTS 2.0）	`VOLC_TTS_API_KEY`
Tavily	联网调研搜索	`TAVILY_API_KEY`
抖音	视频发布	Cookie via auto-douyin

English

Pipeline Overview

Idea → Script → Review → Assets → Code → Review → Render → Publish
 │      │        │        │       │       │        │        │
 │   /video-  /script-   TTS    /remotion- /video-  render  /douyin-
 │   script   review     cover   video     review           publish

Step	Command	What it does
1. Script	`/video-script <idea>`	Research topic online, generate storyboard script with shot-by-shot visual direction, voiceover text, and on-screen text
2. Review	`/script-review <slug>`	Fact-check every claim against authoritative sources, assess quality, auto-fix issues
3. Voiceover	`/voiceover <slug>`	Generate TTS audio via Volcano Ark API, output per-shot MP3 files + Whisper-aligned subtitle timing
4. Cover	`/video-cover <slug>`	Generate 4:3 landscape + 3:4 portrait covers using AI image generation + text overlay
5. Code	`/remotion-video <slug>`	Convert storyboard script into Remotion React compositions with layout primitives, animations, transitions, and audio
6. Review	`/video-review <slug>`	Audit Remotion code for Douyin compliance, animation quality, script consistency; auto-fix issues
7. Render	`npx remotion render`	Export MP4 video
8. Publish	`/douyin-publish <slug>`	Auto-extract metadata from script and publish to Douyin

Quick Start

Prerequisites:

Claude Code CLI (the skills run inside Claude Code)
Node.js 18+ and pnpm
Python 3.7+ (for cover text overlay)

Setup:

git clone https://github.com/yasinshaw/media.git
cd media
cp .env.example .env  # Fill in API keys below
cd remotion && pnpm install

API Keys:

Copy .env.example to .env and fill in the following keys:

Key	Purpose	How to get it
`IMAGE_API_KEY` + `IMAGE_API_BASE_URL` + `IMAGE_MODEL`	AI image generation (any OpenAI-compatible provider)	Bltcy relay / OpenAI / Gemini relay — see `.env.example`
`VOLC_TTS_API_KEY`	TTS voiceover synthesis	Volcano TTS Console → API Key Management
`TAVILY_API_KEY`	Web search for research	Tavily → Sign up → API Keys

Only configure the keys for features you need. For example, script generation alone only requires TAVILY_API_KEY.

Create Your First Video:

# 1. Generate a script from an idea
/video-script DeepSeek V4 发布了，推理能力怎么样

# 2. Fact-check the script
/script-review deepseek-v4

# 3. Generate voiceover audio
/voiceover deepseek-v4

# 4. Generate video covers
/video-cover deepseek-v4

# 5. Convert to Remotion code (preview in Studio)
/remotion-video deepseek-v4

# 6. Review code + render MP4
/video-review deepseek-v4 --render

# 7. Publish to Douyin
/douyin-publish deepseek-v4

Skills

`/video-script` — Storyboard Script Generation

Transforms a video idea into a complete storyboard script:

Research phase: Searches official docs, GitHub repos, and authoritative sources
Angle selection: Proposes 3 creative angles (or skips if the idea is clear)
Script generation: Hook → Pain Point → Core Content → CTA
Timing: Chinese speech rate (4-5 chars/sec), duration derived from word count
Visual types: remotion, real-shot, fixed-image, ai-image, ai-video, ai-background
Audio annotations: BGM style/tempo, sound effects per shot

Output: script.md, voiceover.md, research.md

`/script-review` — Fact-Checking & Quality Review

Rigorous review with source verification:

Fact-checks every claim (statistics, product info, dates) against authoritative sources
Validates timing (word count vs. shot duration)
Auto-fixes: unverified numbers, casual language, hyperbole
Asks user before: major factual corrections, narrative changes

`/voiceover` — TTS Audio Generation

Generates voiceover via Volcano Ark TTS API:

Default speaker: Liu Fei 2.0 (mature male, suitable for tech content)
Per-shot split files for flexible editing
Whisper-aligned subtitle timing manifest
Full merge file for quick preview

`/video-cover` — Cover Image Generation

AI-generated covers for multiple platforms:

4:3 landscape (Bilibili, Xigua Video)
3:4 portrait (Douyin, Kuaishou, WeChat)
AI generates pure background → Python adds text overlay with recommended colors

`/remotion-video` — Remotion Code Generation

Converts storyboard scripts into production-ready React video code:

Layout primitives: CenteredStack, HubLayout, TwoColumnCompare, TimelineFlow
Animations: Spring presets, interpolate, typewriter, word highlighting
Transitions: TransitionSeries with fade/slide/wipe/flip/clock-wipe
Audio: Per-shot voiceover, BGM, sound effects
AI backgrounds: configurable OpenAI-compatible image API for cinematic shot backgrounds
Visual effects: Light leaks, motion blur, starburst, noise grain, Lottie, charts

Output: Shot components + composition + Remotion Studio preview

`/video-review` — Code Review & Render

Automated quality gate for Remotion compositions:

Remotion best practices compliance (no CSS transitions, staticFile() for audio, etc.)
Douyin format compliance (1080x1920, safe areas, subtitle positioning)
Script-to-code consistency (shot count, timing, subtitle text)
Visual alignment (SVG coordinates, absolute/flex mixing, z-index)
Animation smoothness (no boolean opacity jumps)
Auto-fixes most issues; supports --render flag for one-step review + render

`/douyin-publish` — Douyin Publishing

Auto-publishes to Douyin:

Extracts title, tags, and cover from project files
Generates topic tags based on content type
Uses auto-douyin for API-based publishing

Project Structure

media/
├── projects/                    # One directory per video / 每个视频一个目录
│   ├── 2026-04-28-ying-shi-ju-feng-ai-workflow/
│   │   ├── script.md            # Storyboard script / 分镜脚本
│   │   ├── voiceover.md         # Extracted voiceover text / 口播文案
│   │   ├── review.md            # Fact-check report / 审核报告
│   │   ├── research.md          # Research notes & sources / 调研文档
│   │   ├── assets/
│   │   │   ├── images/          # AI-generated covers / AI 封面
│   │   │   └── audio/           # TTS audio files / 配音文件
│   │   └── output/              # Rendered MP4 / 渲染输出
│   └── 2026-04-25-deepseek-v4/
│       ├── script.md            # Storyboard script / 分镜脚本
│       ├── voiceover.md         # Extracted voiceover text / 口播文案
│       ├── review.md            # Fact-check report / 审核报告
│       ├── research.md          # Research notes & sources / 调研文档
│       ├── assets/
│       │   ├── footage/         # Camera recordings / 人物录制素材
│       │   ├── images/          # AI-generated covers / AI 封面
│       │   └── audio/           # TTS audio files / 配音文件
│       └── output/              # Rendered MP4 / 渲染输出
│
├── remotion/                    # Remotion project / Remotion 项目
│   ├── src/
│   │   ├── components/          # Reusable components / 可复用组件
│   │   │   ├── CenteredStack   #   Layout: vertical stack / 垂直堆叠
│   │   │   ├── HubLayout       #   Layout: center + surrounding / 中心+周围
│   │   │   ├── TwoColumnCompare#   Layout: side-by-side / 左右对比
│   │   │   ├── TimelineFlow    #   Layout: sequential flow / 顺序流程
│   │   │   ├── SafeArea        #   Douyin safe area / 抖音安全区域
│   │   │   ├── Subtitle        #   Static subtitle / 静态字幕
│   │   │   ├── ProgressiveSubtitle # Word-by-word subtitle / 逐词字幕
│   │   │   ├── BGMAudio        #   Background music / 背景音乐
│   │   │   └── SFXLayer        #   Sound effects / 音效
│   │   ├── projects/<slug>/    # Per-video compositions / 按项目组合
│   │   │   ├── composition.tsx
│   │   │   └── shots/
│   │   └── root.tsx
│   └── public/
│       ├── audio/              # BGM, SFX, voiceover / 音乐、音效、配音
│       └── images/             # AI-generated backgrounds / AI 背景
│
├── scripts/
│   ├── add_cover_text.py       # Cover text overlay / 封面文字叠加
│   └── generate_audio_assets.py # Synthetic BGM/SFX generator / 合成音频生成
│
└── .claude/skills/             # Claude Code skills / Claude Code 技能
    ├── video-script/           #   Script generation / 脚本生成
    ├── script-review/          #   Fact-checking / 事实核查
    ├── voiceover-tts/          #   TTS audio generation / TTS 配音
    ├── video-cover/            #   Cover image generation / 封面生成
    ├── remotion-video/         #   Remotion code generation / Remotion 代码
    ├── video-review/           #   Code review + render / 代码审核+渲染
    └── douyin-publish/         #   Douyin publishing / 抖音发布

Remotion Components

Component	Purpose
`CenteredStack`	Vertical content stack with safe area + subtitle
`HubLayout`	Center node + 8 surrounding positions with SVG connections
`TwoColumnCompare`	Side-by-side comparison (vertical/horizontal)
`TimelineFlow`	Sequential items with badges and connectors
`SafeArea`	Douyin safe area wrapper (120px top, 200px bottom, 40px sides)
`ProgressiveSubtitle`	Word-by-word subtitle with Whisper-aligned timing
`BGMAudio`	Background music with style mapping and volume control
`SFXLayer`	Sound effects layer (whoosh, impact, text-pop, outro)

External Services

Service	Purpose	Env Variable
OpenAI-compatible image API	Image generation (relay / OpenAI / Gemini relay, configurable)	`IMAGE_API_KEY` / `IMAGE_API_BASE_URL` / `IMAGE_MODEL`
Volcano TTS	Voiceover synthesis (TTS 2.0)	`VOLC_TTS_API_KEY`
Tavily	Web search for research	`TAVILY_API_KEY`
Douyin	Video publishing	Cookie via auto-douyin

License

MIT

Name		Name	Last commit message	Last commit date
Latest commit History 84 Commits
.claude		.claude
docs/superpowers		docs/superpowers
projects		projects
remotion		remotion
scripts		scripts
.env.example		.env.example
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
CLAUDE.md		CLAUDE.md
README.md		README.md
skills-lock.json		skills-lock.json
video-review.skill		video-review.skill
自媒体sop.md		自媒体sop.md

Folders and files

Latest commit

History

Repository files navigation

Media Toolkit

Showcase / 作品展示

影视飓风 AI 工作流 (2026-04-28)

DeepSeek V4 (2026-04-25)

中文

流水线概览

快速开始

技能详情

/video-script — 分镜脚本生成

/script-review — 事实核查与质量审核

/voiceover — TTS 配音生成

/video-cover — 封面图生成

/remotion-video — Remotion 代码生成

/video-review — 代码审核与渲染

/douyin-publish — 抖音发布

Remotion 组件

外部服务

English

Pipeline Overview

Quick Start

Skills

/video-script — Storyboard Script Generation

/script-review — Fact-Checking & Quality Review

/voiceover — TTS Audio Generation

/video-cover — Cover Image Generation

/remotion-video — Remotion Code Generation

/video-review — Code Review & Render

/douyin-publish — Douyin Publishing

Project Structure

Remotion Components

External Services

License

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages