make-prompt-seedance2 skill
A skill that turns an idea into ready Seedance 2.0 storyboards for Jimeng: it reads references, then picks one of five methods
- 764 ★ on the repository
- Last commit
Minimal access
Why this level
- The skill writes only prompt text and runs nothing. One route makes clips that resemble a filmed person, and they should be labelled as AI-made.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add skills/make-prompt-seedance2Detects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewMinimal accessThe tool runs nothing and makes no network calls.
- The skill writes only prompt text and runs nothing. One route makes clips that resemble a filmed person, and they should be labelled as AI-made.
This is third-party code. Review the repository files before installing.
What it does
The `seedance-video-prompt` skill prepares prompts for Seedance 2.0, a ByteDance video model available in the Jimeng, Xiaoyunque and Doubao apps. The agent builds a storyboard and timeline so the text can be copied into the service. First it analyses uploaded references: product, background, character, lettering and style. Then it picks one of five mutually exclusive methods: editing existing video, a realistic phone-shot style clip, a sales clip, a TVC brand advert or a general creative scenario. For the chosen method the agent reads the right file from `references/` and builds the prompt by its formula. Hard rules are fixed in the skill: one shot carries one action, references to materials are written as `@图片N`, `@视频N` and `@音频N`, and the agent does not invent a slogan.
Who it is for. Short-video authors and online-shop owners who generate clips in Seedance and want ready prompts for different tasks.
Good fit when
- A storyboard for a 15-second product advert is needed in one request from a photo of the packaging.
- An existing video has to be extended or its plot changed, with the prompt referring to it as `@视频1`.
- A brand advert is made from five shots of three seconds each with a final logo.
- A prompt has to be built for the vertical Douyin format with a hook in the first three seconds.
Not a fit when
- A clip from another video model is needed: the reference formats and limits are built for Seedance 2.0.
- The video has to show a real person's face, which the model does not support for such materials.
- The clip gives a false impression of a live review without a note that it is AI-made.
Example request
Prepare a Seedance 2.0 prompt: a 15-second vertical advert for a thermo mug from the photo in `@图片1`, five shots, a final logo frame with no invented slogan.Limitations
The skill only writes prompts, while videos are generated in Jimeng, Xiaoyunque or Doubao, which run from China and may be unavailable or need a separate account. A prompt may reference up to 9 images, 3 videos and 3 audio files. Seedance does not accept real faces in materials, and for well-known characters the skill replaces names with a description of appearance to avoid refusal. The repository and method texts are in Chinese. The repository states no license.
How to disable. Delete the `seedance-video-prompt` folder from the agent's skills directory. Installation leaves no other traces.
FAQ
How do I install the skill in Codex or Claude Code?
Copy `skills/seedance-video-prompt/` together with its `references/` folder into the agent's skills directory. The README has no install command.
Which methods does the skill contain?
Five: video editing, a realistic clip, sales (two versions: for TikTok and for Douyin), TVC and a general scenario. The agent chooses by priority and the clip's goal.
What if one hero has to stay consistent across shots?
Provide a reference image of the character. The model does not accept real faces, so the skill reminds about this limit.
Where is the video generated from the prompt?
In Jimeng (jimeng.jianying.com), Xiaoyunque or Doubao. The skill only prepares the text and a list of materials to upload.
How long can the clip be?
15 seconds by default with a supported range of 4–15 seconds. The default format is 16:9 or 9:16.
Discussion
No questions or reviews yet.
Related
SkillsA skill that teaches an agent to write prompts for ByteDance's Jimeng Seedance 2.0 video model: @ references, camera language, timing, templates
The skill breaks a sales video into a storyboard and writes a prompt pack for the Seedance video model in Jimeng.
A dialogue that turns an idea into a storyboard and a prompt for the Seedance 2.0 video model in Jimeng, with six templates and ten examples
A skill that turns an idea, theme or novel excerpt into a 30-second multi-shot prompt for the Seedance 2.5 video model.
A skill that turns a video idea into Seedance 2.5 settings, reference roles and one paste-ready prompt for the video model.
A skill that improves Seedance prompts, plans storyboards and generates video through Atlas Cloud, plus a library of ready prompts
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Repository stars | Main difference |
|---|---|---|---|
| make-prompt-seedance2 (this card) | no | 764 | |
| Seedance2 Storyboard Generator | no | 2,842 | the same author: a separate storyboard generator; here one skill with five methods and reference analysis |
| Ecom Video Seedance Prompt | no | 125 | a narrow skill for product clips only; here also TVC, video editing and a general scenario |
README in short
The README explains what Seedance 2.0 is and where to use it and gives a table of ready templates for advertising, product display, a short plot, a fight, an interior tour and a sound track. It ends with the author's sibling projects for presenters and video cloning.
README badge
Original SKILL.md
--- name: seedance-video-prompt description: 生成 AI 短视频（Seedance 2.0 / 即梦 / 小云雀）分镜提示词的专家。当用户想生成任何形式的短视频提示词、视频脚本、分镜脚本时使用——包括但不限于：剧情/叙事视频、打斗/动作视频、风景旅拍、短剧、伪纪录片、家装空间漫游、UI 产品宣传片、口播视频、音乐卡点、品牌广告片 TVC、产品展示、电商带货视频（抖音/TikTok）、真人感种草/探店视频、视频延长或改剧情。只要用户提到"生成视频提示词"、"帮我写分镜"、"做个短视频"、"带货视频"、"产品广告"等就应触发。 --- # Seedance 2.0 短视频提示词生成 你的职责：先理解用户上传的参考素材（如有），把用户的视频想法路由到正确的提示词方法论，收集必要信息，生成可直接复制到 Seedance 2.0（即梦 jimeng.jianying.com / 小云雀 / 豆包）使用的完整提示词。 本 skill 内含 5 套互斥的方法论，存放于 `references/`。**不要凭记忆套模板——先路由，再读取对应的参考文档，严格按文档中的公式、模板和输出格式生成。** ## 第零步：理解用户上传的参考素材（如有） 用户经常会上传参考图（产品图、角色图、场景图、分镜图，或本地文件/附件/图片 URL）。**只要用户上传了图片，必须先逐张查看并分析内容，再进入路由和生成——严禁跳过图片直接凭文字描述编造画面。** 用 Read 工具查看本地图片文件；远程 URL 图片用图像分析工具查看。每张图按四个维度提取信息： 1. **产品/主体**：是什么、类型、颜色（尽量精确到色感）、形状、材质、关键细节 2. **背景环境**：商超/居家/户外/影棚/实验室等，光线特征 3. **人物形象**：性别、年龄段、穿搭、表情动作（如有） 4. **文字与风格**：图中的 Logo/卖点文字/价格牌；色彩基调、构图、光影、质感调性 图片分析结果的三种用途： - **填充路线变量**：从图片直接读出产品类型、品牌调性、目标场景，减少向用户追问（e.g. 图是白底产品图 → 带货路线；图是带 Logo 的品牌视觉 → TVC 路线） - **精确描述核心主体**：提示词中的产品/角色描述必须与参考图实际内容一致（颜色、材质、细节照实写），这是提升 Seedance 还原度的关键——描述与参考图冲突时模型会生成扭曲结果 - **素材引用规划**：确定每张图在提示词中的角色（`@图片1 作为首帧`、`@图片2 产品外观`、`@图片3 角色形象` 等），并在最终输出的"素材建议"中告诉用户上传顺序与用途对应关系 带货路线特别注意：`references/ecommerce-playbook.md` 的"图片内容分析与广告词语提取方法"章节支持根据图片内容反推最合适的带货玩法，优先按它决策。 ## 第一步：路由决策 按以下顺序判断（命中即停，优先级从上到下）： | 优先级 | 路线 | 触发信号 | 参考文档 | |---|---|---|---| | 1 | **视频编辑类** | 用户已有视频，要延长、续拍、修改剧情、音乐卡点、口播对口型 | `references/general-storyboard.md`（模板五/六/十一/十二） | | 2 | **TikTok 真人感带货** | 要求"看不出是 AI"、"真实感"、"手机随手拍"、探店/种草/口播/氛围类，平台是 TikTok 或强调 UGC 质感 | `references/realperson-video.md` | | 3 | **电商带货** | 以卖货转化为目标：有产品、卖点、投放平台、痛点，要带货视频/商品展示 | 见下方带货二选一 | | 4 | **品牌广告 TVC** | 品牌宣传片、广告大片：有品牌调性、Slogan/Logo 落版需求，追求电影级质感 | `references/tvc-ad.md` | | 5 | **通用创意** | 其余一切：叙事、打斗、旅拍、短剧、战争、伪纪录片、空间漫游、UI 产品动效、长镜头追踪等 | `references/general-storyboard.md` | ### 带货二选一 - **海外 TikTok / 痛点明确**（信任感缺失、售后焦虑、效果难可视化、质感显廉价等）→ `references/ecommerce-playbook.md`：先按"痛点诊断表"从 11 种玩法中选型，再按其输出格式生成。