seedance-auto-generate

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Seedance 2.0 Auto-Generate via Playwright

通过 Playwright 自动生成 Seedance 2.0

This skill automates the full Higgsfield Seedance 2.0 video creation flow using Playwright browser automation. It uploads an image, fills in a prompt, and clicks Generate — hands-free.

Source: github.com/beshuaxian/higgsfield-seedance2-jineng

此 Skill 使用 Playwright 浏览器自动化完成 Higgsfield Seedance 2.0 视频创作的完整流程。它会上传图片、填写提示词并点击 Generate——全程无需手动操作。

来源：github.com/beshuaxian/higgsfield-seedance2-jineng

Available Sub-Skills (Prompt Generators)

可用的子 Skill（提示词生成器）

This auto-generate skill is the automation engine. To craft the best possible prompt for your video, first invoke the matching style sub-skill before running this automation. Each sub-skill turns Claude into a specialized prompt engineer for that video style — generating large, detailed, paste-ready prompts with powerful 2-second hooks that stop the scroll.

这个自动生成 Skill 是自动化引擎。若要为视频编写尽可能出色的提示词，请先调用与风格匹配的子 Skill。每个子 Skill 都会让 Claude 成为该视频风格的专属提示词工程师，生成篇幅长、细节丰富、可直接粘贴的提示词，并利用强有力的 2 秒钩子吸引用户停留、停止滚动。

How to Use Sub-Skills

子 Skill 使用方法

Pick the sub-skill that matches your video style from the table below

Invoke it with

/01-cinematic

/07-ecommerce-ad

/15-real-estate

, etc.

The sub-skill generates a production-grade Seedance 2.0 prompt (15-25 lines, with timeline, camera, lighting, sound, and 2-second hook)
Then invoke
```
/seedance-auto-generate
```
to automatically submit the prompt + image to Higgsfield

选择与视频风格匹配的子 Skill

使用

/01-cinematic

、

/07-ecommerce-ad

、

/15-real-estate

等命令调用该 Skill

子 Skill 会生成一条制作级 Seedance 2.0 提示词（15–25 行，包含时间线、镜头、光线、声音和 2 秒钩子）
然后调用
```
/seedance-auto-generate
```
，将提示词和图片自动提交至 Higgsfield

Creative Styles

创意风格

Slash Command	Skill Name	When to Use
`/01-cinematic`	Cinematic Film	Film quality — dramatic lighting, camera language, depth of field, anamorphic, noir, epic
`/02-3d-cgi`	3D CGI	3D rendered — Pixar, Unreal Engine, photorealistic, isometric, ray tracing, volumetric
`/03-cartoon`	Cartoon & Animation	2D animation — cel-shaded, hand-drawn, flat vector, watercolor, rubber hose, motion graphics
`/04-comic-to-video`	Comic to Video	Animate comics — manga pages, webtoons, storyboards, sequential art, graphic novels
`/05-fight-scenes`	Fight Scenes & Action	Combat — martial arts, sword fights, chase scenes, superhero battles, action choreography
`/08-anime-action`	Anime	Japanese animation — shonen, seinen, mecha, magical girl, isekai, anime openings

斜杠命令	Skill 名称	使用场景
`/01-cinematic`	Cinematic Film	电影质感——戏剧性灯光、镜头语言、景深、变形宽银幕、黑色电影、史诗感
`/02-3d-cgi`	3D CGI	3D 渲染——Pixar、Unreal Engine、照片级真实、等距视角、光线追踪、体积效果
`/03-cartoon`	Cartoon & Animation	2D 动画——赛璐璐渲染、手绘、扁平矢量、水彩、橡皮管动画、动态图形
`/04-comic-to-video`	Comic to Video	漫画动画——漫画页、条漫、分镜、连续艺术、图像小说
`/05-fight-scenes`	Fight Scenes & Action	战斗——武术、剑斗、追逐戏、超级英雄大战、动作编排
`/08-anime-action`	Anime	日式动画——少年漫画、青年漫画、机甲、魔法少女、异世界、动漫片头

Commercial & Marketing

商业与营销

Slash Command	Skill Name	When to Use
`/06-motion-design-ad`	Motion Design Ad	Software/SaaS — product launches, feature showcases, app promos, tech demos, UI animation
`/07-ecommerce-ad`	E-Commerce Ad	Product ads — fashion, beauty, electronics, food, Amazon/Shopify/TikTok Shop, unboxing
`/09-product-360`	Product 360	Turntable — multi-angle showcase, product reveal, hero shots, beauty shots, product spin
`/11-social-hook`	Social Hook	Viral content — TikTok, Instagram Reels, YouTube Shorts, scroll-stopping hooks
`/12-brand-story`	Brand Story	Brand narrative — origin stories, mission videos, company culture, founder stories

斜杠命令	Skill 名称	使用场景
`/06-motion-design-ad`	Motion Design Ad	软件/SaaS——产品发布、功能展示、应用推广、技术演示、UI 动画
`/07-ecommerce-ad`	E-Commerce Ad	产品广告——时尚、美妆、电子产品、食品、Amazon/Shopify/TikTok Shop、开箱展示
`/09-product-360`	Product 360	转台展示——多角度展示、产品揭幕、英雄镜头、美妆镜头、产品旋转
`/11-social-hook`	Social Hook	病毒式内容——TikTok、Instagram Reels、YouTube Shorts、吸引用户停留的钩子
`/12-brand-story`	Brand Story	品牌叙事——起源故事、使命视频、企业文化、创始人故事

Industry-Specific

行业专用

Slash Command	Skill Name	When to Use
`/10-music-video`	Music Video	Beat-synced — performance, lyric video, music visualization, concert visuals, album art
`/13-fashion-lookbook`	Fashion Lookbook	Fashion — lookbooks, model walks, outfit showcases, runway clips, streetwear, campaigns
`/14-food-beverage`	Food & Beverage	Food — restaurant promos, recipe content, food ASMR, menu showcases, appetite appeal
`/15-real-estate`	Real Estate	Property — house tours, architecture, interior design, listings, virtual tours, renovation

斜杠命令	Skill 名称	使用场景
`/10-music-video`	Music Video	节拍同步——表演、歌词视频、音乐可视化、演唱会视觉、专辑封面
`/13-fashion-lookbook`	Fashion Lookbook	时尚——lookbook、模特走秀、穿搭展示、跑道片段、街头服饰、广告宣传片
`/14-food-beverage`	Food & Beverage	美食——餐厅推广、菜谱内容、美食 ASMR、菜单展示、增强食欲
`/15-real-estate`	Real Estate	房产——房屋导览、建筑、室内设计、房源列表、虚拟看房、装修展示

What Each Sub-Skill Generates

每个子 Skill 生成的内容

Every sub-skill provides:

2-Second Hook Framework — 10-12 attention-grabbing opener patterns
Timeline Segmentation — beat-by-beat breakdown up to 15s
Camera Movement Encyclopedia — 15-20+ techniques with exact phrasing
Lighting & Atmosphere — setups that communicate mood and quality
Sound Design — ambient, foley, music, silence guidance
Material Reference Strategy —
```
@image1
```
```
@video1
```
```
@audio1
```
best practices
Platform Optimization — TikTok, Instagram, YouTube, etc.
5+ Large Example Prompts — 15-25 lines each, production-quality

Each skill also has a Chinese version at
skills/XX-name/zh-CN/SKILL.md

每个子 Skill 都会提供：

2 秒钩子框架——10–12 种能够吸引注意力的开场模式
时间线分段——最长 15 秒的逐拍拆解
镜头运动大全——20 多种技术及精确表述方式
灯光与氛围——能够传达情绪和品质的布光方案
声音设计——环境声、拟音、音乐和静默的使用建议
素材引用策略——
```
@image1
```
、
```
@video1
```
、
```
@audio1
```
的最佳实践
平台优化——适配 TikTok、Instagram、YouTube 等平台
5 个以上大型示例提示词——每个提示词 15–25 行，达到制作级质量

每个 Skill 还有中文版本，路径为
skills/XX-name/zh-CN/SKILL.md

Seedance 2.0 Specs

Seedance 2.0 规格

Input	Format	Limit
Image	jpeg, png, webp, bmp, tiff, gif	Up to 9, each < 30MB
Video	mp4, mov	Up to 3, each < 50MB, 2-15s total
Audio	mp3, wav	Up to 3, each < 15MB, up to 15s total
Text	Natural language prompt	—
Combined	—	Max 12 files total
Output	Video	4-15s, 720p, with sound

Reference materials in prompts:

@image1

@video1

@audio1

输入	格式	限制
图片	jpeg、png、webp、bmp、tiff、gif	最多 9 张，每张小于 30MB
视频	mp4、mov	最多 3 个，每个小于 50MB，总时长 2–15 秒
音频	mp3、wav	最多 3 个，每个小于 15MB，总时长最多 15 秒
文本	自然语言提示词	—
合计	—	最多共 12 个文件
输出	视频	4–15 秒、720p、包含声音

提示词中的参考素材：

@image1

、

@video1

、

@audio1

Prerequisites

前置条件

User must be logged in to higgsfield.ai in the Playwright browser session
Playwright MCP plugin must be enabled
Image file must be accessible within the project directory (copy from Downloads if needed)

用户必须已在 Playwright 浏览器会话中登录 higgsfield.ai
必须启用 Playwright MCP 插件
图片文件必须位于项目目录内可访问的位置（如有必要，可从 Downloads 复制）

Input Requirements

输入要求

The user provides:

An image file (path on disk —
```
.png
```
,
```
.jpg
```
,
```
.webp
```
, etc.)
A description or prompt for the video (or invoke a sub-skill first to generate one)

Optional overrides (defaults shown):

Duration:
```
8s
```
(options: 5s, 8s, 10s)
Ratio:
```
Auto
```
(options: Auto, 16:9, 9:16, 1:1)
Resolution:
```
720p
```
(options: 720p, 1080p)

用户提供：

一个图片文件（磁盘路径，例如
```
.png
```
、
```
.jpg
```
、
```
.webp
```
等）
一段视频描述或提示词（也可以先调用子 Skill 生成提示词）

可选覆盖项（括号内为默认值）：

时长：
```
8s
```
（可选：5s、8s、10s）
比例：
```
Auto
```
（可选：Auto、16:9、9:16、1:1）
分辨率：
```
720p
```
（可选：720p、1080p）

Automation Flow

自动化流程

Follow these steps exactly using Playwright MCP tools. Load each tool with

ToolSearch

before first use.

使用 Playwright MCP 工具严格按以下步骤操作。每次首次使用某个工具前，先通过

ToolSearch

加载该工具。

Step 0: Prepare the image

第 0 步：准备图片

If the image is outside the project directory (e.g. in

C:\Users\User\Downloads\

), copy it into the project root first:

cp "<source_path>" "C:/Users/User/higgz/<filename>"

The Playwright file upload tool only allows files within the project root.

如果图片位于项目目录之外（例如

C:\Users\User\Downloads\

），请先将其复制到项目根目录：

cp "<source_path>" "C:/Users/User/higgz/<filename>"

Playwright 文件上传工具仅允许上传项目根目录内的文件。

Step 1: Navigate to Seedance 2.0 creation page

第 1 步：打开 Seedance 2.0 创建页面

browser_navigate → url: "https://higgsfield.ai/create/video?model=seedance_2_0"

The

?model=seedance_2_0

query parameter pre-selects the Seedance 2.0 model automatically.

Wait for the page to load, then take a snapshot to confirm:

The page title contains "Create AI Videos"
The Model button shows "Seedance 2.0"

If a cookie consent dialog appears, close it by clicking the "Close" button (x).

browser_navigate → url: "https://higgsfield.ai/create/video?model=seedance_2_0"

?model=seedance_2_0

查询参数会自动预选 Seedance 2.0 模型。

等待页面加载完成，然后获取快照以确认：

页面标题包含 "Create AI Videos"
Model 按钮显示 "Seedance 2.0"

如果出现 Cookie 同意对话框，请点击 "Close" 按钮（x）将其关闭。

Step 2: Open the upload dialog

第 2 步：打开上传对话框

Take a snapshot and find the upload area. It contains text like "Upload media" and "Image, Video or Audio". Click the upload area element.

This opens a dialog with tabs: "Uploads", "Image Generations", "Video Generations", "Elements", "Liked".

获取快照并找到上传区域。其中包含类似 "Upload media" 和 "Image, Video or Audio" 的文本。点击上传区域元素。

这会打开一个对话框，其中包含以下标签页："Uploads"、"Image Generations"、"Video Generations"、"Elements"、"Liked"。

Step 3: Trigger the file chooser

第 3 步：触发文件选择器

Inside the dialog, find and click the "Upload media" button (the first clickable item in the dialog grid — it has an upload icon and text "Upload media" with subtext "Protected content is not allowed").

This triggers a file chooser modal. The tool result will say:

[File chooser]: can be handled by browser_file_upload

在对话框内找到并点击 "Upload media" 按钮（它是对话框网格中的第一个可点击项——带有上传图标和 "Upload media" 文本，副文本为 "Protected content is not allowed"）。

这会触发一个文件选择器模态框。工具结果将显示：

[File chooser]: can be handled by browser_file_upload

Step 4: Upload the image file

第 4 步：上传图片文件

browser_file_upload → paths: ["C:\\Users\\User\\higgz\\<filename>"]

Use the absolute Windows path with double backslashes. The file must be within the Playwright allowed roots (

C:\Users\User\higgz

browser_file_upload → paths: ["C:\\Users\\User\\higgz\\<filename>"]

使用带双反斜杠的 Windows 绝对路径。文件必须位于 Playwright 允许的根目录（

C:\Users\User\higgz

）内。

Step 5: Wait for upload to process

第 5 步：等待上传处理

browser_wait_for → time: 5

Wait 5 seconds for the image to upload and appear in the asset grid.

browser_wait_for → time: 5

等待 5 秒，让图片完成上传并显示在素材网格中。

Step 6: Select the uploaded image

第 6 步：选择已上传的图片

Take a snapshot of the dialog. The newly uploaded image appears as the first item after the "Upload media" button in the grid. It will be a

generic [cursor=pointer]

element. Click it.

Important: The newly uploaded image is always the first grid item right after the upload button. Look for the element immediately following the upload button in the dialog's grid.

获取对话框的快照。新上传的图片会显示为网格中紧跟在 "Upload media" 按钮之后的第一个项目。它会是一个

generic [cursor=pointer]

元素。点击该元素。

重要：新上传的图片始终是上传按钮之后网格中的第一个项目。请查找对话框网格中紧接上传按钮的元素。

Step 7: Close the dialog

第 7 步：关闭对话框

browser_press_key → key: "Escape"

The dialog closes. The image is now loaded into the form — the upload area is replaced by an image preview.

browser_press_key → key: "Escape"

对话框随即关闭。图片已加载到表单中——上传区域会替换为图片预览。

Step 8: Verify image is loaded

第 8 步：确认图片已加载

Take a snapshot. Confirm:

The upload area now shows an image preview (a
```
figure
```
with
```
img "media is loading"
```
or similar)
The prompt textbox is visible (labeled "Prompt" with placeholder "Describe your scene in detail")
The Model still shows "Seedance 2.0"

获取快照并确认：

上传区域现在显示图片预览（一个包含
```
img "media is loading"
```
或类似内容的
```
figure
```
）
提示词文本框可见（标签为 "Prompt"，占位符为 "Describe your scene in detail"）
Model 仍显示 "Seedance 2.0"

Step 9: Enter the prompt

第 9 步：输入提示词

Click the prompt textbox, then type the prompt:

browser_click → ref: <textbox_ref>, element: "Prompt textbox"
browser_type → ref: <textbox_ref>, text: "<the video prompt>"

The prompt should be crafted by one of the 15 sub-skills listed above. If the user hasn't specified a style, pick the best-matching sub-skill and use it to generate the prompt first.

点击提示词文本框，然后输入提示词：

browser_click → ref: <textbox_ref>, element: "Prompt textbox"
browser_type → ref: <textbox_ref>, text: "<the video prompt>"

提示词应使用上述 15 个子 Skill 之一进行编写。如果用户未指定风格，请选择最匹配的子 Skill，并先使用它生成提示词。

Step 10: Adjust settings (optional)

第 10 步：调整设置（可选）

If the user wants non-default settings:

Duration: Click the "Duration" button, then select the desired option
Ratio: Click the "Auto Ratio" button, then select (16:9, 9:16, 1:1)
Resolution: Click the "720p Resolution" button, then select (720p, 1080p)

如果用户需要使用非默认设置：

时长：点击 "Duration" 按钮，然后选择所需选项
比例：点击 "Auto Ratio" 按钮，然后选择 16:9、9:16 或 1:1
分辨率：点击 "720p Resolution" 按钮，然后选择 720p 或 1080p

Step 11: Generate

第 11 步：生成

Take a screenshot to show the user the final form state before generating.

ASK THE USER FOR CONFIRMATION before clicking Generate — this costs credits.

browser_click → ref: <generate_button_ref>, element: "Generate button"

The Generate button shows the credit cost (e.g., "Generate 48 32"). After clicking, the video enters the generation queue.

点击 Generate 之前，先截取屏幕截图，向用户展示最终的表单状态。

点击 Generate 之前，务必请求用户确认——此操作会消耗积分。

browser_click → ref: <generate_button_ref>, element: "Generate button"

Generate 按钮会显示积分消耗，例如 "Generate 48 32"。点击后，视频将进入生成队列。

Step 12: Monitor generation

第 12 步：监控生成过程

After clicking Generate:

The page may redirect or show a progress indicator
Click the "History" tab to see generation status
Wait and periodically check until the video is ready
Once complete, the video will appear in the History tab

点击 Generate 后：

页面可能会重定向或显示进度指示器
点击 "History" 标签页查看生成状态
等待并定期检查，直到视频准备就绪
完成后，视频会显示在 History 标签页中

Key Element Patterns (for snapshot navigation)

关键元素模式（用于快照导航）

These are the typical element patterns found on the page. Refs change between sessions — always take a fresh snapshot before interacting.

Element	How to find it
Upload area	`generic [cursor=pointer]` containing text "Upload media" and "Image, Video or Audio"
Upload button in dialog	First item in `dialog > generic > generic` grid, contains "Upload media" text
Newly uploaded image	First `generic [cursor=pointer]` after the upload button in dialog grid
Prompt textbox	`textbox` inside `generic` with label "Prompt"
Model button	`button "Model"` containing "Seedance 2.0"
Duration button	`button "Duration"`
Ratio button	`button "Auto Ratio"` or `button "16:9 Ratio"`
Resolution button	`button "720p Resolution"`
Generate button	`button "Generate ..."` with credit cost numbers

以下是页面上常见的元素模式。Refs 会在不同会话中发生变化——交互前始终应获取最新快照。

元素	查找方式
上传区域	包含 "Upload media" 和 "Image, Video or Audio" 文本的 `generic [cursor=pointer]`
对话框中的上传按钮	`dialog > generic > generic` 网格中的第一项，包含 "Upload media" 文本
新上传的图片	对话框网格中紧接上传按钮之后的第一个 `generic [cursor=pointer]`
提示词文本框	标签为 "Prompt" 的 `generic` 内的 `textbox`
Model 按钮	包含 "Seedance 2.0" 的 `button "Model"`
时长按钮	`button "Duration"`
比例按钮	`button "Auto Ratio"` 或 `button "16:9 Ratio"`
分辨率按钮	`button "720p Resolution"`
Generate 按钮	带有积分消耗数字的 `button "Generate ..."`

Error Handling

错误处理

Not logged in: If the page shows a login prompt instead of the creation form, ask the user to log in manually (
```
! open https://higgsfield.ai
```
in the CLI)
File chooser not triggered: Retry clicking the upload media button; sometimes a second click is needed
Image not appearing in grid: Wait longer (up to 10s) and re-snapshot
Dialog won't close: Try clicking outside the dialog area, or press Escape again
Insufficient credits: The Generate button shows the cost — warn the user before clicking

未登录：如果页面显示登录提示而不是创建表单，请让用户手动登录（在 CLI 中执行
```
! open https://higgsfield.ai
```
）
未触发文件选择器：重试点击上传媒体按钮；有时需要再次点击
图片未显示在网格中：等待更长时间（最多 10 秒），然后重新获取快照
对话框无法关闭：尝试点击对话框区域外部，或再次按 Escape
积分不足：Generate 按钮会显示所需积分——点击前应提醒用户

Example Workflows

示例工作流

Example 1: Real Estate Video (end-to-end)

示例 1：房地产视频（端到端）

User: "Create a video for this house photo at C:\Users\User\Downloads\house.webp"

Step 1 → Invoke /15-real-estate with the image to craft a cinematic real estate prompt
Step 2 → Copy image to project: cp Downloads/house.webp higgz/house.webp
Step 3 → Run this automation flow with the image + generated prompt
Step 4 → Confirm with user before clicking Generate
Step 5 → Monitor until video is complete

User: "Create a video for this house photo at C:\Users\User\Downloads\house.webp"

Step 1 → Invoke /15-real-estate with the image to craft a cinematic real estate prompt
Step 2 → Copy image to project: cp Downloads/house.webp higgz/house.webp
Step 3 → Run this automation flow with the image + generated prompt
Step 4 → Confirm with user before clicking Generate
Step 5 → Monitor until video is complete

Example 2: Product Ad

示例 2：产品广告

User: "Make a TikTok ad for my sneakers" + [sneaker photo]

Step 1 → Invoke /07-ecommerce-ad to craft a product showcase prompt
Step 2 → Copy image, run automation, set ratio to 9:16 for TikTok vertical
Step 3 → Confirm and generate

User: "Make a TikTok ad for my sneakers" + [sneaker photo]

Step 1 → Invoke /07-ecommerce-ad to craft a product showcase prompt
Step 2 → Copy image, run automation, set ratio to 9:16 for TikTok vertical
Step 3 → Confirm and generate

Example 3: Anime-Style Fight Scene

示例 3：动漫风格战斗场景

User: "Turn this character art into an anime fight scene"

Step 1 → Invoke /08-anime-action to craft an anime action prompt
Step 2 → Copy image, run automation
Step 3 → Confirm and generate

User: "Turn this character art into an anime fight scene"

Step 1 → Invoke /08-anime-action to craft an anime action prompt
Step 2 → Copy image, run automation
Step 3 → Confirm and generate

Example 4: Quick Generate (no sub-skill)

示例 4：快速生成（不使用子 Skill）

User: "Just generate a video from this image, cinematic style"

Step 1 → Write a concise cinematic prompt directly (or auto-invoke /01-cinematic)
Step 2 → Copy image, run automation
Step 3 → Confirm and generate

User: "Just generate a video from this image, cinematic style"

Step 1 → Write a concise cinematic prompt directly (or auto-invoke /01-cinematic)
Step 2 → Copy image, run automation
Step 3 → Confirm and generate