com um clique
ai-video-studio
一键直出国际范大牌广告片 - 4个Skills串联,6步出片,从静态图到30秒成片
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Menu
一键直出国际范大牌广告片 - 4个Skills串联,6步出片,从静态图到30秒成片
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
Direct browser control via CDP. Use when the user wants to automate, scrape, test, or interact with web pages. Connects to the user's already-running Chrome.
Convert any input into a modern editorial black-and-white line-art character illustration system and produce PNG image-generation outputs by default. Use when the user asks to turn text, images, brand names, product ideas, UI concepts, or campaigns into minimalist geometric character illustrations, editorial layouts, website hero visuals, mobile UI mockups, packaging, magazine spreads, or PNG visuals in the specified monochrome line-art plus pastel accent style.
Debug red border/rectangle overlays appearing on Chinese text in HyperFrames HTML on macOS Chrome — caused by images with baked-in borders, not CSS.
Extract frames from a reference video at precise timestamps to analyze opening animation timing and visual style before building a HyperFrames composition.
Typeset professional documents: resumes, one-pagers, white papers, letters, portfolios, slide decks. Warm parchment, ink-blue accent, serif-led hierarchy. CN uses TsangerJinKai02, EN uses Charter, JA uses YuMincho (best-effort). Triggers on "做 PDF / 排版 / 一页纸 / 白皮书 / 作品集 / 简历 / PPT / slides", or "build me a resume / make a one-pager / design a slide deck / turn this into a PDF / make this presentable".
Generate professional SVG logos and high-end showcase images. Use when the user wants to: (1) Create a logo or icon for their product/brand, (2) Generate logo design concepts based on product information, (3) Create professional logo showcase presentations with multiple background styles, (4) Export logos in various formats (SVG, PNG), or (5) Iterate on logo designs with different visual styles. Supports geometric patterns, dot matrix designs, line systems, and mixed compositions. Generates showcase images using Nano Banana (Gemini image generation) with 12 professional background styles.
| name | ai-video-studio |
| description | 一键直出国际范大牌广告片 - 4个Skills串联,6步出片,从静态图到30秒成片 |
| metadata | {"version":"2.0","author":"Claude Code Skill System","tags":["video","ai","advertising","remotion","elevenlabs","jimeng"]} |
核心价值:4个 Skills 串联,6步出片,从静态图到 30 秒成片。
AI 做视频的人都知道痛点:
解决方案:用 Skill 系统把 4 个独立工具串成一条自动化流水线。
| 工具 | 用途 |
|---|---|
| Claude | 剧本 + 提示词 + 编排 |
| 即梦 | 图生视频 |
| ElevenLabs | AI 配音 + 音效 |
| Remotion | React 声明式合成 |
① 剧本 → Claude 生成镜头脚本
② 提示词 → 每镜头的画面描述 + 运镜指令
③ 静态帧 → AI 生图 (Midjourney/即梦/SD)
④ 图生视频 → 即梦 API (5秒/镜头)
⑤ 配音+音效 → ElevenLabs TTS + Sound Effects
⑥ 合成 → Remotion 声明式编排 → 渲染输出
bash skills/ai-video-studio/scripts/init.sh my-ad-project
# 提供品牌信息,Claude 输出镜头脚本
bash skills/ai-video-studio/modules/script-writing/generate.sh \
--brand "Equation" \
--product "香薰蜡烛" \
--style "Aesop-inspired minimalism"
bash skills/ai-video-studio/modules/prompt-engineering/generate.sh \
--script ./scripts/script.json \
--output ./prompts/
bash skills/ai-video-studio/modules/image-to-video/generate.sh \
--images ./images/ \
--prompts ./prompts/ \
--output ./videos/
bash skills/ai-video-studio/modules/voiceover/generate.sh \
--script ./scripts/script.json \
--output ./audio/
bash skills/ai-video-studio/modules/sound-effects/generate.sh \
--script ./scripts/script.json \
--output ./audio/sfx/
cd remotion/
npm install
# 预览
npx remotion studio
# 渲染
npx remotion render EquationAd out/equation-ad.mp4
写「Aesop 风格」比写「克制、极简、优雅」有效 10 倍。
生 8 张选 5 张比反复修改一张快得多。
对于高端品牌广告来说,英文更有「大牌感」。
这是黄金比例。
工具可替换,改一行配置就能换工具:
# config/providers.yaml
voiceover:
default: elevenlabs
providers:
elevenlabs: { strengths: [英文优秀, 音效API] }
minimax: { strengths: [全球第一, 中文极佳, 6秒克隆] }
image_to_video:
default: jimeng
providers:
jimeng: { strengths: [运镜控制好, 国内访问] }
runway: { strengths: [质量高, Gen-3 Alpha] }
seedance: { strengths: [质量极高, 2.0版本], status: planned }
ad-commercial - 品牌广告片mv-music-video - MV 剪辑product-showcase - 产品展示social-short - 社媒短视频[场景] + [产品细节] + [光线] + [色调hex] + [技术参数] + [品牌锚点]
即梦是字节跳动的产品,对中文运镜指令的理解更准确。
ai-video-studio/
├── SKILL.md # 入口文件
├── config/
│ └── providers.yaml # Provider 切换
├── workflows/ # 5 种场景工作流
│ ├── ad-commercial.md
│ ├── mv-music-video.md
│ └── ...
├── modules/ # 能力模块
│ ├── script-writing/
│ ├── prompt-engineering/
│ ├── image-to-video/
│ ├── voiceover/
│ ├── sound-effects/
│ └── composition/
├── templates/ # 可运行的脚本模板
│ └── remotion/
└── knowledge/ # 领域知识
├── motion-vocabulary.md
└── troubleshooting.md
| 问题 | 原因 | 解决方案 |
|---|---|---|
| interpolate [0,0] 报错 | fadeIn=0 导致区间 [0,0] | 加判断:fadeIn > 0 ? interpolate(...) : 1 |
| 渲染时符号链接失败 | Remotion 不支持 symlink | 直接复制素材到 public/ |
| 音画不同步 | 配音时长 ≠ 镜头时长 | 先生成配音,根据实际时长调整 durationInFrames |
| 即梦运镜没效果 | 提示词太复杂 | 一个镜头一个主要动作 |
| 中文配音「土」 | 不是技术问题,是定位问题 | 高端品牌用英文 |
| 旁白远超视频长度 | 中文TTS约1:1实时率,原始文案时长无法预估 | 先生成各段TTS获取真实时长,重新计算每段开始时间,再用ffmpeg adelay精确对齐 |
| 多段TTS拼接位置错位 | 脚本中预设的start_time只是目标,实际语音时长由语速决定 | 不要用脚本预设时间,直接用ffmpeg adelay把每段音频放到计算出的绝对时间点 |
这套系统的本质不是「AI 帮你做视频」,而是「AI 帮你搭建一条可复用的视频生产线」。
Skills 就该这么用——不是一个万能的大 Skill,而是多个小 Skill 的灵活组合。像乐高一样拼接,每次组合都可能产生新的玩法。
下一步:等 Seedance 2.0 API 开放后,替换掉即梦,画面质量会再上一个台阶。配置已经预留好了。