Seedance 2.0 Video Gen

Seedance 2.0 AI 视频生成技能。根据用户描述（文字/图片/视频/音频）， 自动生成 Seedance 2.0 API 调用参数，支持文生视频、图生视频、参考视频三种模式。 支持 @素材名 语法控制角色、镜头、动作、声音。 触发词：生成视频、制作视频、AI视频、图生视频、文生视频、视频生成

Security Scan

The skill's behavior (calling the EvoLink/Seedance API using an API key) matches its stated purpose, but the package metadata omits the required SEEDANCE_API_KEY and the skill lacks a verifiable source/homepage — this metadata/instruction mismatch and missing provenance are concerning.

Purpose & Capability

The name/description and SKILL.md consistently describe a video-generation helper that builds prompts and produces EvoLink/Seedance API call parameters — that capability aligns with the declared models and API endpoints. However, the registry metadata lists no required environment variables while the SKILL.md explicitly requires an API key (SEEDANCE_API_KEY), which is an inconsistency between what the skill says it needs and what the registry declares.

Instruction Scope

The runtime instructions are scoped to constructing prompts/JSON and showing curl/fetch/Python examples for POSTing to https://api.evolink.ai/v1/videos/generations and polling tasks. The instructions do not direct reading unrelated system files or exfiltrating data beyond calls to the third‑party API. They do reference using an Authorization header with $SEEDANCE_API_KEY (not declared in registry).

Install Mechanism

This is an instruction-only skill with no install spec and no code files — nothing is written to disk and no packages are installed. That is the lowest-risk install mechanism.

Credentials

The SKILL.md requires an API key (SEEDANCE_API_KEY) for EvoLink/Seedance, which is appropriate for a remote video-generation service. But the registry metadata lists zero required env vars and no primary credential — the omission is a mismatch that can obscure the fact the skill will attempt to use an external secret. No other unrelated credentials are requested.

Persistence & Privilege

The skill does not request always: true and is user-invocable; it does not attempt to modify other skills or agent-wide config. Autonomous invocation is allowed by default but not combined with other high privileges here.

What to consider before installing

This skill appears to do what it claims (prepare parameters for Seedance/EvoLink video generation) and uses a single API key. Before installing: 1) Confirm the skill will need SEEDANCE_API_KEY (the SKILL.md requires it) and update registry settings or only provide the key when you intend to call the external API. 2) Verify the service endpoints and vendor (evolink.ai / seedance2api.app) independently — the skill has no homepage or source link in the registry. 3) Be cautious uploading sensitive images/audio/video: assets will be sent to a third‑party API and generated video links expire in 24h. 4) Prefer creating an API key with minimal permissions and monitor usage/quotas. 5) If you need higher assurance, ask the publisher for the source repository or a signed homepage and confirm the registry metadata is corrected to declare required env vars.

Seedance 2.0 视频生成技能

版本：1.0.0 API 来源：EvoLink.ai 文档：https://seedance2api.app

---

核心能力

1. 理解需求 → 选择合适的模型（文生视频/图生视频/参考视频）
2. 构建 Prompt → 使用 @素材名 语法精确控制内容
3. 生成 API 参数 → 输出符合 EvoLink API 规范的 JSON
4. 提供完整调用示例 → 附 curl / fetch / Python 代码

---

模型选择指南

用户需求	推荐模型
纯文字描述生成视频	seedance-2.0-text-to-video
一张图动画化	seedance-2.0-image-to-video
多模态参考（图+视频+音频组合）	seedance-2.0-reference-to-video
快速迭代测试	*-fast-* 系列（更快）

---

API 基础信息

接口地址：

POST https://api.evolink.ai/v1/videos/generations
GET  https://api.evolink.ai/v1/tasks/{task_id}

认证： Authorization: Bearer $SEEDANCE_API_KEY

环境变量： SEEDANCE_API_KEY（需用户在 EvoLink.ai 注册获取）

---

工作流程

Step 1：理解用户需求

必问信息（如未提供则推断）：

• 内容主题（人物/产品/场景/抽象概念）
• 是否需要参考素材（图/视频/音频）
• 视频时长（4-15秒，默认5秒）
• 画面比例（16:9 / 9:16 / 1:1 / 4:3 / 3:4 / 21:9）
• 是否需要音效

判断模型类型：

• 只有文字 → Text-to-Video
• 有1-2张参考图 → Image-to-Video
• 有图+视频/音频/多图 → Reference-to-Video

Step 2：构建 Prompt（@素材名语法）

基础规则：

@image1  = 第1张参考图（人物外观/场景/产品）
@video1  = 第1个参考视频（镜头运动/角色动作）
@video2  = 第2个参考视频（辅助镜头/动作）
@audio1  = 第1个参考音频（背景音乐/音效）

常用写法：

@image1 as the first frame  # 指定首帧
Reference camera movement from @video1  # 参考镜头
Reference character action from @video1  # 参考动作
Background BGM references @audio1  # 背景音乐
Extend @video1 by 5s  # 视频延长

Step 3：生成 API 调用参数

Text-to-Video：

{
  "model": "seedance-2.0-text-to-video",
  "prompt": "中文或英文 prompt",
  "duration": 5,
  "quality": "720p",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

Image-to-Video（1张图）：

{
  "model": "seedance-2.0-image-to-video",
  "prompt": "@image1 as the first frame, ...",
  "image_urls": ["https://..."],
  "duration": 5,
  "quality": "720p",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

Reference-to-Video（多模态）：

{
  "model": "seedance-2.0-reference-to-video",
  "prompt": "@image1 as the first frame, @video1 as camera reference, ...",
  "image_urls": ["https://..."],
  "video_urls": ["https://..."],
  "audio_urls": ["https://..."],
  "duration": 5,
  "quality": "720p",
  "aspect_ratio": "16:9",
  "generate_audio": true
}

Step 4：输出完整调用示例

提供 curl / Python / JavaScript 三种调用方式。

---

输入规格

类型	格式	数量	单文件限制
图片	jpeg, png, webp, bmp, tiff, gif	≤9张	<30MB
视频	mp4, mov	≤3个	<50MB
音频	mp3, wav	≤3个	<15MB

总上限： ≤12个文件（图+视频+音频）

---

输出规格

参数	可选值
时长	4-15秒，或 -1（智能时长）
分辨率	480p / 720p
比例	16:9 / 9:16 / 1:1 / 4:3 / 3:4 / 21:9 / adaptive
音效	generate_audio: true（含音效/配乐）

视频链接有效期： 24小时

---

价格参考

分辨率	标准价格
480p	4.63 credits/秒
720p	10.00 credits/秒

（音频生成不额外收费）

---

注意事项

1. 合规：不支持写实真人脸，建议使用插画/AI虚拟角色
2. 镜头控制：镜头运动需通过 @video1 参考提供，纯文字描述镜头效果不稳定
3. 视频延长：Extend 时 prompt 需描述原视频最后一帧的状态作为起点
4. One-Shot 写法：结尾必须加 "No scene cuts throughout, one continuous shot."
5. 长视频分段：10秒以上建议用时间轴分段：0-3s: ... / 3-6s: ...

---

相关资源

• API 文档：https://seedance2api.app
• 获取 API Key：https://evolink.ai/signup
• Prompt 案例库：prompts/
• 官方用例参考：cases/