Seedance Storyboard

把剧本/小说/文字脚本/分镜文案，转换成 Seedance（即梦 / 豆包 Doubao Seedance）文生视频用的分镜脚本——带「视觉基调 + 色彩与调影 + 情绪基调」全局头，再加按秒切分的「分镜1、分镜2…」逐镜运镜与动作描写。只要用户给一段剧情/动作/场景文字，并提到要做成视频、AI 视频、Seedance、即梦、Sora、可灵、Veo、文生视频、视频脚本、分镜、运镜、storyboard、shot list，或想把小说/桥段/广告脚本变成可生成的视频提示词，就用这个 skill。即使用户只说「帮我把这段改成视频脚本」「拆成分镜」「写成 AI 视频提示词」也要触发。

6 stars 2 forksWriting & Content

Install

skills CLI (Claude Code, Cursor, Codex and more)

npx skills add zcx960/seedance-storyboard --skill seedance-storyboard

Manual install

git clone --depth 1 https://github.com/zcx960/seedance-storyboard.git /tmp/seedance-storyboard
mkdir -p ~/.claude/skills
cp -r /tmp/seedance-storyboard/. ~/.claude/skills/seedance-storyboard

Personal skills live in ~/.claude/skills. Use .claude/skills in a project folder to share a skill with your team.

From the SKILL.md

# Seedance 分镜脚本生成器

把一段**剧情文字**（剧本、小说桥段、动作描述、广告脚本、一句话创意）转写成 **Seedance / 即梦** 文生视频可直接喂入的**分镜脚本**：先定全局影像风格，再按秒拆成逐镜的运镜 + 动作描写。

输出风格对标专业的影视化分镜文案——精确到秒、运镜明确、动作连贯、画面可拍。

## 为什么是这个结构

Seedance 的提示词遵循 **主体 + 动作 + 场景 + 光影 + 镜头语言 + 风格** 的骨架，模型很擅长在清晰输入上做扩展。单条视频通常 **5s 或 10s @ 24fps**；多镜头要靠**显式的切换/衔接提示**（如「无缝切至」「镜头切换」），切镜后**要重新交代新场景**。模型**不响应负向提示词**，所以「不要…」要改写成正向描述；脸部/人体/角色一致性这类**稳定性约束**写进风格头里能显著减少崩坏。

整套脚本因此分两层：**全局头**（一次性定调，所有分镜共享）+ **分镜时间轴**（逐镜的镜头与动作）。

## 工作流程

1. **读懂剧情**：先把输入读成「谁、在哪、做什么、什么情绪、什么节奏」。先不急着写格式。
2. **判断是单镜片段还是多场景长剧本**：
   - **单场景 / 一段连续动作**（一镜到底或几次切换，总时长 ≤10s）→ 输出**一条** Seedance 分镜脚本。这是默认情形，也是质量最高的用法。
   - **多场景 / 长剧本**（出现明显的时间或地点跳转、多段对白、章节）→ 先按场景切分，**每个场景产出一条独立的 Seedance 分镜脚本**，并在每条前加 `### 场景 N · <一句话场景名>` 标题。每条仍各自 ≤10s。不要把跨场景的内容塞进同一条 10s 脚本里。
3. **定全局影像风格**（见下方模板的三个头）：从剧情的类型、年代、情绪反推镜头器材、色彩调性、情绪节奏。剧本没写的，按类型合理补全（科幻冷调、古装暖调、悬疑低照度…），但不要无中生有地改变剧情设定。
4. **拆分镜时间轴**：把这段动作拆成若干 beat，每个 beat 一个分镜，标注**时间区间**，让区间累加 ≈ 目标总时长（默认 ~10s；用户指定了就按指定）。每个分镜先给**景别 + 运镜**，再写**主体 + 动作 + 细节**。相邻分镜要能接上（动作的尾接下一镜的头）。
5. **自检**：时间轴是否连续不重叠？每镜是否都有景别和运镜？是否全程同一主体/服装一致？有没有把「不要 X」写成了负向提示（要改成正向）？

## 输出模板

严格用下面这个结构（单场景就只出一份；多场景每个场景出一份，前面加场景标题）。中文输出，术语对齐影视分镜习惯。

Read the full skill on GitHub

Details

Skill name: seedance-storyboard
Repository: zcx960/seedance-storyboard
License: MIT
Repo updated: 26 June 2026

FAQ

What does the Seedance Storyboard skill do?

How do I install the Seedance Storyboard skill?

skills CLI (Claude Code, Cursor, Codex and more): npx skills add zcx960/seedance-storyboard --skill seedance-storyboard. Skills go in ~/.claude/skills (personal) or .claude/skills (project), and Claude loads them automatically when a task matches the description.

Is Seedance Storyboard popular?

It lives in zcx960/seedance-storyboard, which has 6 GitHub stars and 2 forks.

Related skills

Sharextract

wuaishare/sharextract

Extract normalized content from public share URLs, RSS/Atom feeds, timed-text/subtitle documents, and web pages using a protocol-first fallback ladder. Use for public AI chat shares, social/media links, RSS/Atom feeds, WebVTT/SRT/TTML captions, articles, oEmbed pages, public JSON endpoints, or when an agent needs the highest-fidelity public content without bypassing authentication, CAPTCHAs, paywalls, or access controls.

Writing & Content

Content Repurposer

Mahanaicoach/content-repurposer

Turn one short-form video (an Instagram Reel or a TikTok) into platform-correct versions for Instagram, TikTok, and YouTube — then auto-schedule TikTok and YouTube through the user's own Zernio account. It transcribes the video and OCRs the frames to find the call-to-action, then TRANSLATES that CTA per platform: Instagram keeps "comment <word> and I'll send you the link" (the DM mechanic), while TikTok becomes "link in bio" and YouTube becomes "link in description" — in the on-screen text, the voiceover, AND the caption. If the CTA only lives in the description, it rewrites the description pe

Writing & Content

Humanizer Ru

unicodef1wn/humanizer-russian-language

Removes signs of AI generation from Russian-language blog content and social media posts (Telegram, vc.ru, Habr and similar) while keeping a lively, readable style. Universal and voice-neutral — works for long articles, short posts, notes, news breakdowns and opinion pieces. Must be used when the user says "перепиши более живо" (rewrite it more lively), "сделай менее ИИ-шно" (make it less AI-ish), "звучит как ChatGPT" (sounds like ChatGPT), "слишком сухо/шаблонно" (too dry / formulaic), "выглядит как LinkedIn-пост" (looks like a LinkedIn post), "очисти от воды" (cut the fluff). Also apply proa

Writing & Content

Storyboard

Penny777btc/storyboard

分镜（镜头序列）设计判断库：把口播脚本/创意拆解成「一镜一镜」的镜头序列——景别序列、机位角度、镜头切换、动作衔接、时长节奏、镜头清单。不是工具，而是给「写脚本之后、拍摄/录屏之前的分镜规划」与「选片定景别、粗剪排序」提供分镜判断依据。当需要把文字变成画面、决定一段话拆几个镜头/每个镜头是什么景别和机位、镜头之间怎么接、或质检「这段为什么接得不顺」时调用。管「镜头怎么排序」；画面内构图交给 [[shot-language]]，机位走位交给 [[camera-blocking]]，文字图形交给 [[layout-design]]。

Writing & Content

Ill Communication

alexconner-79/ill-communication

Write and review human-sounding copy, with deep UX writing craft. Use this skill for EVERY task that produces prose of any kind (chat responses, documents, presentations, emails, case studies, reports, blog posts) AND whenever the user asks to review, humanize, de-AI, edit, or improve any text or UX copy (buttons, error messages, empty states, microcopy, onboarding, notifications). It applies even when the user does not mention style, tone, or punctuation. If the task involves writing, rewriting, or judging any sentence, this skill applies.

Writing & Content

Humanize Writing

geyadawang-boop/humanize-writing

创作、续写和修订中文或英文文本，改善空洞、重复和模板化表达。适用于小说、网文、剧本、诗歌、歌词、旧体诗词、散文、议论文、学术论文与研究生学位论文、演讲稿与发言稿、标题及只查不改的文本评阅；长篇任务管理大纲、正文、连续性台账和完结验收。检测分数仅供编辑参考，不鉴定作者身份或保证过检。

Writing & Content

Browse the most popular Claude skills, all writing & content skills, or write your own with the skills generator.