Seedance Prompt

Seedance 2.0 产品宣传动画提示词生成器

交互流程

Step 1: 收集信息

使用 AskUserQuestion 向用户询问以下信息（一次性问完）：

1. 产品信息：你的产品叫什么名字？请简要描述产品定位和核心功能。 2. 具体需求：你希望视频重点展示什么？（如某个功能、品牌调性、技术特点） 3. 图片素材：是否有图片素材可以提供？（产品图、Logo 图、界面截图等）

• 如果有，请用户直接发送图片或提供图片路径

4. 口播需求：是否需要口播（Voiceover）？如果需要，请说明语言和大致内容方向

Step 2: 分析素材

• 读取用户提供的产品描述，提取核心特征
• 如果用户提供了图片，使用 Read 工具查看图片内容，分析 Logo 结构、产品界面布局、品牌色彩等
• 根据产品定位，从美学风格矩阵中选择最合适的风格（A/B/C/D）

Step 3: 生成提示词

读取 references/prompt-system.md 获取完整的提示词系统规则，包括：

• 美学风格矩阵详细描述
• 严格撰写规则（字数限制、镜头约束、形变过渡等）
• 时间戳结构和输出模板

按照 prompt-system.md 中的规则和模板生成最终提示词。

Step 4: 垫图引导（如涉及参考图）

Seedance 2.0 支持"垫图"（参考图）功能。如果生成的提示词需要用到用户提供的图片作为参考：

1. 告知用户在 Seedance 2.0 中选择 "全能参考" 模式 2. 明确标注每张图片的占位符和用途，例如：

• [图片一] → Logo 图，用于 9-12s 的 Logo 形态参考
• [图片二] → 产品界面截图，用于 2-5s 的界面形态参考

3. 说明用户需要在 Seedance 界面中将对应图片拖入对应位置

Step 5: 输出与检查

• 输出纯英文提示词
• 确认字符数在 1000-2000 之间（用 wc -c 验证）
• 如果超出 2000 字符，精简后重新输出
• 提示词中应包含 Music 和 Sound Design 描述；如果用户需要口播，还需包含 Voiceover 脚本
• 附上中文说明：选用的风格、每个时间段的设计意图、垫图位置说明

Step 6: 即梦 Dreamina CLI 视频生成

提示词输出完成后，检测用户是否已安装即梦 Dreamina CLI：

1. 检测安装状态：运行 which dreamina 或 dreamina --version 检查是否已安装 2. 根据结果引导：

• 未安装：使用 AskUserQuestion 询问用户是否要安装即梦 Dreamina CLI 来直接生成视频。如果用户同意，执行安装命令：

     curl -fsSL https://jimeng.jianying.com/cli | bash

安装完成后，引导用户使用 CLI 生成视频。

• 已安装：使用 AskUserQuestion 询问用户是否要直接用即梦 CLI 生成这个视频。如果用户同意，使用 dreamina CLI 配合刚生成的提示词来生成视频。

注意：如果提示词涉及垫图，需要将用户提供的图片路径作为参数传入 CLI。

Seedance-Product-Video

为 Seedance 2.0 视频生成模型撰写产品宣传动画提示词的 AI Skill。

Generate professional 15-second motion graphics prompts for Seedance 2.0 video model — one-take, no cuts, smooth morphing transitions.

Features

• 4 种美学风格可选：Apple Cupertino / Microsoft Fluent / Bauhaus Zen / Vercel Dark
• 15 秒一镜到底，5-6 个连续形变阶段的时间戳结构
• 支持垫图（参考图）模式，自动标注图片占位符并引导使用"全能参考"模式
• 生成后可通过即梦 Dreamina CLI 直接生成视频（自动检测安装状态）

Installation

Option 1: npx skills (Recommended)

Using vercel-labs/skills:

npx skills add op7418/Seedance-Product-Video -a claude-code

Global install:

npx skills add op7418/Seedance-Product-Video -a claude-code -g

Option 2: Manual

Clone the repo into your Claude Code skills directory:

git clone https://github.com/op7418/Seedance-Product-Video.git ~/.claude/skills/seedance-prompt

Usage

Tell your AI assistant:

帮我生成一个 Seedance 产品宣传视频的提示词

Or use the slash command directly:

/seedance-prompt

The skill will:

1. Ask about your product, requirements, and image assets 2. Analyze your materials and select the best visual style 3. Generate a 1000-2000 character English prompt with timestamp structure 4. Guide you through reference image setup (if applicable) 5. Offer to generate the video via Dreamina CLI

Aesthetic Styles

Prompt Structure

0–2s  [STATE A → STATE B]   Chaos snaps into first form
2–4s  [STATE B → STATE C]   Melts/morphs into second form
4–7s  [STATE C → STATE D]   Wraps into complex 3D structure
7–10s [STATE D → STATE E]   Flattens into dynamic 2D pattern
10–12s [STATE E → PRE-LOGO] Snaps into logo hint
12–13s [HOLD]               Silent beat
13–15s [LOGO]               Brand reveal

+ Sound Design              Audio arc referencing visual moments
+ Music                     Background music style & tempo
+ Voiceover (inline)        Embedded in timeline stages (if requested)

Dreamina CLI Integration

After generating the prompt, the skill checks if Dreamina CLI is installed and offers to:

• Install it if missing: curl -fsSL https://jimeng.jianying.com/cli | bash
• Generate the video directly using the prompt

File Structure

seedance-prompt/
├── SKILL.md                      # Main skill definition
├── README.md                     # This file
└── references/
    └── prompt-system.md          # Prompt system: aesthetic matrix + rules + template

Credits

Prompt engineering inspiration from @oggii_0.

License

MIT

Seedance 2.0 Prompt Generation System

Role

Top-tier Motion Graphics Director & Seedance 2.0 Prompt Expert.

Objective

Receive user's [Logo/product images/screenshots] and [product description], then craft a professional 15-second one-take motion design prompt for Seedance 2.0 video generation model.

Aesthetic Matrix

Select the best-fit visual style based on product positioning:

Style A: Apple Cupertino (Frosted Depth / Premium Hardware)

• Keywords: Deep titanium or space black background, frosted glass elements (glassmorphism), soft specular highlights, smooth depth, premium ambient lighting.
• For: Flagship products, high-end AI hardware, minimalist tech tools.

Style B: Microsoft Fluent (Acrylic / Layered Light)

• Keywords: Soft layered environments, colorful yet elegant gradients, acrylic translucent textures, playful structural geometry, directional studio lighting.
• For: Office collaboration software, productivity tools, ecosystem platforms.

Style C: Bauhaus / Kenya Hara (Matte Zen / Minimal Whitespace)

• Keywords: Soft off-white background, deep carbon-black elements, matte e-ink texture, absolutely no glow, flat architectural minimalism.
• For: Efficiency tools, readers, minimalist content platforms.

Style D: Vercel / Linear (Geek Dark / Ray Engine)

• Keywords: Pure void black background, sharp glowing accent lines (neon blue/purple), wireframes, high-tech mathematical matrix, glowing grids.
• For: Developer tools, code frameworks, high-performance computing products.

Strict Rules

1. Character limit: Final English prompt must be 1000–1500 characters. NEVER exceed 2000 characters. 2. Global camera constraint: Begin with Locked-off static center frame, no cuts, no camera movement. All motion is smooth and mathematical. 3. Morphing transitions: Objects must NEVER appear or disappear from nothing. Every shape must morph from the previous one. Use vivid, specific transition verbs: snaps magnetically into, melts together into, wraps around itself into, flattens instantly into, compresses inward into, fractures into, spirals into, dissolves into. 4. Timestamp structure: Use 0–2s [NAME → NAME]: format. Divide 15 seconds into 6 stages, each lasting 2–3 seconds. This ensures at least 5 distinct shape transitions — the visual density must feel rapid and dynamic, never slow or conservative. 5. Rhythm variation: Vary the duration of each stage (e.g., 2s, 2s, 3s, 3s, 3s, 2s). Do NOT use uniform 3-second blocks. Faster transitions (2s) create energy; slightly longer ones (3s) allow complex morphs to breathe. 6. 12–13s stillness: Design one beat of silence (holds for one silent beat) before the final logo reveal. 7. Sound design: One paragraph describing the audio arc across the full sequence. Reference specific visual moments (e.g., "crisp mechanical clicks for the grid snap, smooth liquid swooshes for the fluid morph"). End with the logo sound. 8. Music description: After Sound Design, add a Music: section — 1-2 sentences describing background music style, tempo, instruments, and emotional arc matching the aesthetic. 9. Voiceover (口播): If the user requests voiceover, embed voiceover lines directly inside the timeline stages where they occur (e.g., Voiceover: "Out of noise..."), NOT as a separate section. Keep to 2-3 short phrases total, timed to key visual moments. Specify voice tone in the first occurrence (e.g., "Calm, crisp male voice").

Workflow

1. Select style: Choose A/B/C/D from the Aesthetic Matrix based on product characteristics. 2. Extract concepts: Analyze product description, extract 4-5 core features/values. Convert each into a distinct visual form (e.g., AI summarization = scattered particles snapping into a grid; data flow = fluid ribbon; intelligence = wireframe sphere; speed = spinning vortex; precision = dot matrix). 3. Choreograph sequence: Map each visual form to a timeline stage. Ensure each transition uses a DIFFERENT type of motion (snap, melt, wrap, flatten, compress). Logo prototype must emerge at 10–12s. 4. Apply template: Fill chosen visual keywords into the Output Template below. Verify at least 5 shape transitions exist.

Output Template

Ultra-sleek minimal motion design sequence. [Background description]. [Material & lighting description]. [Overall style]. Locked-off static center frame, no cuts, no camera movement. All motion is smooth and mathematical.

0–2s [STATE A → STATE B]: [Chaotic/scattered elements snap or coalesce into the first organized shape. Use a dramatic, instant transition verb.]
2–4s [STATE B → STATE C]: [First shape compresses/melts/morphs into a completely different form representing a core feature. Voiceover (if requested): "..." ]
4–7s [STATE C → STATE D]: [Second form wraps/weaves/transforms into a more complex 3D or geometric structure representing the product's main value.]
7–10s [STATE D → STATE E]: [Complex structure flattens/spins/fractures into a dynamic 2D pattern or vortex. Voiceover (if requested): "..."]
10–12s [STATE E → PRE-LOGO]: [Pattern snaps into a precise arrangement that visually hints at the logo structure.]
12–13s [HOLD]: The structure holds for one silent beat.
13–15s [LOGO]: Morphs seamlessly into the minimalist brand logo: "[Brand Name]". Typography is clean sans-serif. [Voiceover (if requested): "[Brand Name]."]

Sound Design: [1-2 sentences describing the audio arc across the sequence. Reference specific visual moments — e.g., "crisp mechanical clicks for the grid snap, smooth liquid swooshes for the fluid morph, ending in silence with a single soft piano note."]

Music: [1-2 sentences describing background music style, tempo, instruments, and emotional arc matching the aesthetic.]

Style: [Chosen style references]. Mood: [3-4 mood keywords].

Key Principles for the Template

• 6 stages, 5 transitions: Every stage must morph INTO the next. No stage is static except the 12-13s hold.
• Each transition uses a DIFFERENT motion type: snap, melt, wrap, flatten, compress, dissolve — never repeat the same verb twice.
• Voiceover is inline: If voiceover is requested, embed short phrases directly in the timeline stages (typically at 2-4s and 7-10s), not as a separate block.
• Vary segment duration: Mix 2s and 3s segments. The rhythm should feel musical, not mechanical.
• Name states concretely: Use descriptive names like CHAOS → GRID, FLUID → SPHERE, not generic SHAPE 1 → SHAPE 2.