Skip to content
Back to search

seedance-prompt-skill

Seedance 2.0 prompt skill,使用该Skill生成Seedance 2.0 视频提示词

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/songguoxs-seedance-prompt-skill/badge.svg)](/listings/songguoxs-seedance-prompt-skill)

HTML

<a href="/listings/songguoxs-seedance-prompt-skill"><img src="/api/listings/songguoxs-seedance-prompt-skill/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Prerequisites

- [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) installed and authenticated - A Seedance 2.0 (即梦) account at [jimeng.jianying.com](https://jimeng.jianying.com) for using the generated prompts

Steps

1. **Clone the repository**

git clone https://github.com/songguoxs/seedance-prompt-skill.git

2. **Copy the skill into your project's `.claude/skills/` directory**

# Navigate to the project where you want to use the skill
cd /path/to/your/project

# Create the skills directory if it doesn't exist
mkdir -p .claude/skills/seedance

# Copy the skill file
cp /path/to/seedance-prompt-skill/.claude/skills/seedance/SKILL.md .claude/skills/seedance/SKILL.md

3. **Verify installation**

Start Claude Code in your project directory:

claude

Then type `/seedance` or simply describe your video idea. Claude should activate the Seedance prompt skill automatically.

Alternative: Global Installation

To make the skill available across all your projects, place it under your home directory:

mkdir -p ~/.claude/skills/seedance
cp /path/to/seedance-prompt-skill/.claude/skills/seedance/SKILL.md ~/.claude/skills/seedance/SKILL.md

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 333 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. [English](./README.md) | [中文](./README_zh.md)
  2. # Seedance Prompt Skill for Claude Code
  3. A Claude Code custom skill that turns Claude into a professional AI video prompt engineer for ByteDance's **Seedance 2.0** (即梦) video generation platform.
  4. Provide a creative idea in natural language, and this skill will generate structured, production-ready Chinese prompts that you can paste directly into the Seedance 2.0 platform to produce cinematic AI videos.
  5. ## Features
  6. ### Ten Core Prompt Generation Capabilities
  7. | # | Capability | Description |
  8. |---|-----------|-------------|
  9. | 1 | **Pure Text Generation** | Generate videos purely from text descriptions, no reference materials needed |
  10. | 2 | **Consistency Control** | Maintain character/product/scene consistency across shots using reference images |
  11. | 3 | **Camera & Motion Replication** | Replicate camera movements, complex actions, and pacing from reference videos |
  12. | 4 | **Creative Template / VFX Replication** | Reproduce creative transitions, ad templates, and cinematic effects from reference videos |
  13. | 5 | **Story Completion** | Auto-generate storylines from storyboard images or scripts |
  14. | 6 | **Video Extension** | Extend existing videos forward or backward with smooth continuity |
  15. | 7 | **Sound Control** | Voice cloning, dialogue generation, and sound effect design |
  16. | 8 | **One-Take Long Shot** | Generate seamless long takes that flow continuously across scenes |
  17. | 9 | **Video Editing** | Modify existing videos: swap characters, alter plots, add/remove elements |
  18. | 10 | **Music Beat Sync** | Synchronize visual rhythm precisely with music beats |
  19. ### Advanced Techniques
  20. - **Timestamp Storyboarding** — Precise per-second control for 15-second videos (e.g., `0-3s: ...`, `4-8s: ...`)
  21. - **Technical Parameter Specification** — Set aspect ratio, frame rate, color grading, and lens style upfront
  22. - **Negative Prompting** — Explicitly exclude unwanted elements (watermarks, subtitles, etc.)
  23. - **Multi-Segment Stitching** — Automatically split videos longer than 15 seconds into extension-chained segments with continuity notes
  24. ### Multi-Modal Reference System
  25. Supports the Seedance 2.0 `@` reference syntax for combining multiple input modalities:
  26. - **Images** (`@图片1` ~ `@图片9`) — Character references, scene backgrounds, first/last frame control
  27. - **Videos** (`@视频1` ~ `@视频3`) — Camera motion, action, effects, and pacing references
  28. - **Audio** (`@音频1` ~ `@音频3`) — Background music, voice tone, sound effect references
  29. ### Scenario-Specific Prompt Strategies
  30. - **E-commerce / Advertising** — Product showcases, 360-degree spins, 3D exploded views
  31. - **Short Drama / Dialogue** — Scripted scenes with character dialogue, emotion tags, and sound design
  32. - **Xianxia / Fantasy Animation** — Cinematic martial arts, spell effects, transformation sequences
  33. - **Science Education** — 4K medical CGI, transparent anatomy visualizations
  34. - **Music Videos / Beat Sync** — Widescreen compositions, multi-image beat matching
  35. ### Built-in Prompt Asset Libraries
  36. The skill ships with curated vocabularies for:
  37. - **Camera Language** — Shot types, camera movements, angles, pacing, focus, transitions
  38. - **Visual Styles** — Film grains, color palettes, art movements, lighting setups, animation styles
  39. ## Installation
  40. ### Prerequisites
  41. - [Claude Code CLI](https://docs.anthropic.com/en/docs/claude-code) installed and authenticated
  42. - A Seedance 2.0 (即梦) account at [jimeng.jianying.com](https://jimeng.jianying.com) for using the generated prompts
  43. ### Steps
  44. 1. **Clone the repository**
  45. ```bash
  46. git clone https://github.com/songguoxs/seedance-prompt-skill.git
  47. ```
  48. 2. **Copy the skill into your project's `.claude/skills/` directory**
  49. ```bash
  50. # Navigate to the project where you want to use the skill
  51. cd /path/to/your/project
  52. # Create the skills directory if it doesn't exist
  53. mkdir -p .claude/skills/seedance
  54. # Copy the skill file
  55. cp /path/to/seedance-prompt-skill/.claude/skills/seedance/SKILL.md .claude/skills/seedance/SKILL.md
  56. ```
  57. 3. **Verify installation**
  58. Start Claude Code in your project directory:
  59. ```bash
  60. claude
  61. ```
  62. Then type `/seedance` or simply describe your video idea. Claude should activate the Seedance prompt skill automatically.
  63. ### Alternative: Global Installation
  64. To make the skill available across all your projects, place it under your home directory:
  65. ```bash
  66. mkdir -p ~/.claude/skills/seedance
  67. cp /path/to/seedance-prompt-skill/.claude/skills/seedance/SKILL.md ~/.claude/skills/seedance/SKILL.md
  68. ```
  69. ## Usage
  70. ### Quick Start
  71. Start Claude Code and describe what video you want to create:
  72. ```
  73. You: 帮我生成一段赛博朋克风格的城市夜景视频提示词
  74. ```
  75. Claude will ask you a few clarifying questions (duration, aspect ratio, reference materials), then generate 2-3 prompt versions ready to paste into Seedance 2.0.
  76. ### Triggering the Skill
  77. The skill activates automatically when you:
  78. - Use the `/seedance` command
  79. - Mention keywords like: `Seedance`, `即梦`, `视频提示词`, `视频生成`, `AI视频`, `短剧`, `广告视频`, `视频延长`
  80. - Ask to "generate video prompts" or "create Seedance prompts"
  81. ### Interactive Workflow
  82. The skill follows a structured 4-step workflow:
  83. ```
  84. Step 1: You describe your idea → "一段古风武侠打斗"
  85. Step 2: Claude confirms parameters → duration, aspect ratio, references, style preferences
  86. Step 3: Claude generates prompts → 2-3 versions for videos ≤15s, or multi-segment plan for longer videos
  87. Step 4: You refine → adjust timing, camera, dialogue, effects, or style
  88. ```
  89. ## Examples
  90. ### Example 1: Pure Text — Short Action Scene
  91. **Input:**
  92. ```
  93. 帮我写一段仙侠战斗的视频提示词,15秒,横屏
  94. ```
  95. **Output (one of 2-3 versions):**
  96. ```
  97. 15秒仙侠高燃战斗镜头,金红暖色调,0-3秒:低角度特写主角蓝袍衣摆被热浪吹得猎猎飘动,
  98. 双手紧握雷纹巨剑,剑刃赤红电光持续爆闪,地面熔岩翻涌冒泡,远处魔兵嘶吼着冲锋逼近,
  99. 主角低喝"今日,便以这柄剑,镇尔等邪祟!",伴随剑鸣与熔岩咕嘟声;4-8秒:环绕摇镜
  100. 快切,主角旋身挥剑,剑刃撕裂空气迸射红色冲击波,前排魔兵被击飞碎裂成灰烬,伴随剑气
  101. 破空声与魔兵惨嚎;9-12秒:仰拍拉远定格慢放,主角跃起腾空,剑刃凝聚巨型雷光电弧劈
  102. 向魔兵群;13-15秒:缓推特写主角落地收剑的姿态,衣摆余波微动,冷声道"此界之门,
  103. 不容踏越",音效收束为余音震颤与渐弱风声。
  104. ```
  105. ### Example 2: Multi-Modal Reference — Product Ad
  106. **Input:**
  107. ```
  108. 我有一张可乐产品图,想做一个炫酷的产品展示广告
  109. ```
  110. **Output:**
  111. ```
  112. @图片1中的可口可乐饮料,360度高速旋转2圈后,突然停住蓄力分裂成了3个部分进行展示。
  113. 随后分解后的可口可乐饮料罐的上中下三部分快速向内旋转合成,一罐完整的可口可乐饮料,
  114. 3D渲染产品展示特效,动感产品特效展示
  115. 参考素材:
  116. - @图片1:可乐产品正面高清图
  117. ```
  118. ### Example 3: Short Drama with Dialogue
  119. **Input:**
  120. ```
  121. 写一段霸道总裁短剧,15秒,女主撕合同反杀的名场面
  122. ```
  123. **Output:**
  124. ```
  125. 画面(0-5秒):特写女主撕契约镜头,纸屑飘落,总裁单膝跪地伸手阻拦,眼神慌乱,
  126. 女主侧身躲开,嘴角挂着冷漠笑意
  127. 台词1(总裁,卑微慌乱):苏晚!契约还没结束,你不能走!我给你钱,给你地位!
  128. 画面(6-10秒):女主抬脚避开他的手,将撕碎的契约纸扔在他脸上,镜头扫过周围宾客的
  129. 窃窃私语
  130. 台词2(女主,冷漠反杀):契约?顾总,当初是你说,我连给你提鞋都不配,现在求我?晚了!
  131. 画面(11-15秒):总裁僵在原地,脸上沾着纸屑,女主转身昂首离开,红裙裙摆飘动
  132. 音效:华丽又带张力的背景音,契约撕碎的声响,宾客轻微的窃窃私语声
  133. 时长:精准15秒
  134. ```
  135. ### Example 4: Long Video (>15s) — Multi-Segment
  136. **Input:**
  137. ```
  138. 做一段30秒的仙侠出场短片
  139. ```
  140. **Output:**
  141. ```
  142. ## 超长视频提示词(总时长约30秒)
  143. **主题**:剑修云海仙山迎敌
  144. **总段数**:2段
  145. **建议比例**:16:9
  146. ---
  147. ### 第1段(0-15秒)—— 正常生成
  148. **生成时长**:15秒
  149. #### 提示词
  150. 15秒仙侠镜头,0-5秒:俯拍云海翻涌中的仙山全景,镜头缓缓下推穿过云层;
  151. 6-10秒:剑修站在山巅悬崖边,背对镜头,衣袍随风飘动,远处魔气升腾;
  152. 11-15秒:剑修缓缓转身面向镜头,拔剑出鞘,剑刃金光闪烁,目光坚毅低声道
  153. "来了",定格在剑修持剑面向镜头的画面。
  154. #### 衔接点
  155. 本段结尾画面:剑修正面持剑站在山巅,目光前视,远方魔气翻涌。
  156. ---
  157. ### 第2段(15-30秒)—— 视频延长
  158. **操作**:将第1段生成的视频上传为 @视频1
  159. **生成时长**:15秒
  160. #### 提示词
  161. 将@视频1延长15秒。0-5秒:接上段剑修持剑画面,数十只暗影魔兽从远处魔气中
  162. 飞扑而来,剑修纵身跃起迎敌;6-10秒:空中激战,剑气纵横,魔兽被斩成灰烬
  163. 粒子消散,镜头环绕快切;11-15秒:剑修落地收剑,身后爆炸的金色粒子缓缓飘散,
  164. 镜头缓推特写剑修侧脸,音效渐弱。
  165. ```
  166. ### Example 5: One-Take Long Shot with Multi-Image References
  167. **Input:**
  168. ```
  169. 我有5张图,想做一个一镜到底的谍战片追踪镜头
  170. ```
  171. **Output:**
  172. ```
  173. 谍战片风格,@图片1作为首帧画面,镜头正面跟拍穿着红风衣的女特工向前走,
  174. 镜头全景跟随,不断有路人遮挡红衣女子,走到一个拐角处,参考@图片2的拐角建筑,
  175. 固定镜头红衣女子离开画面,走在拐角处消失,一个戴面具的女孩在拐角处躲着恶狠狠
  176. 的盯着她,面具女孩形象参考@图片3。镜头往前摇向红衣女特工,她走进一座豪宅
  177. 消失不见了,豪宅参考@图片4。全程不要切镜头,一镜到底。
  178. 参考素材:
  179. - @图片1:红风衣女特工正面形象(首帧)
  180. - @图片2:旧城区拐角建筑
  181. - @图片3:面具女孩形象
  182. - @图片4:欧式豪宅外观
  183. - @图片5:备用场景(可用于扩展)
  184. ```
  185. ## Seedance 2.0 Platform Specs
  186. | Dimension | Specification |
  187. |-----------|--------------|
  188. | Image Input | jpeg/png/webp/bmp/tiff/gif, up to 9 images, each <30MB |
  189. | Video Input | mp4/mov, up to 3 videos, total 2-15s, each <50MB, 480p-720p |
  190. | Audio Input | mp3/wav, up to 3 files, total ≤15s, each <15MB |
  191. | Text Input | Natural language description |
  192. | File Limit | Max 12 files total (images + videos + audio combined) |
  193. | Duration | 4-15 seconds per generation |
  194. | Sound Output | Built-in sound effects and music |
  195. | Resolution | Up to 2K output |
  196. > **Important:** The platform does not allow uploading images or videos containing realistic human faces. Such materials will be automatically blocked.
  197. ## Tips for Best Results
  198. - **Be specific and visual** — "a woman in a red trench coat walks through rain-soaked neon streets" works much better than "a woman walking"
  199. - **Use timestamp storyboarding** for 13-15 second videos to maintain control over each segment
  200. - **Separate dialogue and sound effects** from visual descriptions for cleaner results
  201. - **Specify negative constraints** at the end (e.g., "no watermarks, no subtitles, no text overlays")
  202. - **Match reference image styles** to your video theme (e.g., use ink-wash style images for Chinese historical themes)
  203. - **For videos over 15 seconds**, let the skill generate multi-segment plans with explicit continuity points between segments
  204. ## Project Structure
  205. ```
  206. .
  207. ├── .claude/
  208. │ └── skills/
  209. │ └── seedance/
  210. │ └── SKILL.md # The skill definition (core file)
  211. ├── .gitignore
  212. └── README.md
  213. ```
  214. The entire skill logic is contained in a single file: `.claude/skills/seedance/SKILL.md`.
  215. ## Contributing
  216. Contributions are welcome. Some areas where help would be appreciated:
  217. - Adding more prompt examples for specific scenarios
  218. - Translating the skill definition to support English prompt generation
  219. - Adding prompt templates for new Seedance features as the platform evolves
  220. - Improving prompt quality based on real-world generation results
  221. To contribute:
  222. 1. Fork this repository
  223. 2. Create a feature branch (`git checkout -b feature/your-feature`)
  224. 3. Commit your changes (`git commit -m "Add your feature"`)
  225. 4. Push to the branch (`git push origin feature/your-feature`)
  226. 5. Open a Pull Request
  227. ## License
  228. MIT
  229. ## Acknowledgments
  230. - [Seedance 2.0](https://jimeng.jianying.com) by ByteDance for the AI video generation platform
  231. - [Claude Code](https://docs.anthropic.com/en/docs/claude-code) by Anthropic for the extensible skill system

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.