PixVerseAI/skills

# PixVerse 技能 用于 [PixVerse CLI](https://www.npmjs.com/package/pixverse) 的 Agent 技能库 — 帮助 AI Agent（Claude Code、Cursor、Codex 等）通过结构化、可组合的工作流生成视频和图像。 ## 这是什么？ PixVerse CLI 是 **[pixverse.ai](https://pixverse.ai) 的无界面版本** — 网站上的所有模型、参数和功能都可以作为 CLI 命令使用，并输出结构化的 JSON。 本代码仓库提供了**技能文件**，用于教导 AI Agent 如何正确使用这些命令：需要传递哪些 flag，哪些模型支持哪些参数，如何将命令串联成 pipeline，以及如何处理错误。 ## 快速开始 ``` # 安装 CLI npm install -g pixverse # Authenticate pixverse auth login # 创建视频 pixverse create video --prompt "A cat astronaut floating in space" --json ``` ## 技能结构 ``` skills/ SKILL.md # Entry point — start here capabilities/ # Individual command skills auth-and-account.md # Authentication & account management create-video.md # Text-to-video, image-to-video, fusion create-and-edit-image.md # Text-to-image, image-to-image modify-video.md # AI content editing (replace subjects, swap outfits, change backgrounds) motion-control.md # Character animation with motion reference video transition.md # Keyframe transition animations post-process-video.md # Extend, upscale, speech, sound prompt-enhance.md # Prompt optimization for V6 video generation task-management.md # Poll and wait for generation tasks asset-management.md # List, download, upload, delete assets saved-folders.md # Organize assets into named folders template.md # Browse and create from effect templates workspace.md # Team workspace management mondo-poster-design.md # Mondo-style poster, book cover, album art design character-design.md # Persistent characters — three-view sheet + cloud asset id reuse item-design.md # Persistent items / props — four-panel orthographic sheet + cloud asset id reuse workflows/ # Multi-step pipeline skills text-to-video-pipeline.md # End-to-end text-to-video image-to-video-pipeline.md # Animate an image into video text-to-image-to-video.md # Generate image then animate it image-editing-pipeline.md # Iterative image editing modify-video-pipeline.md # Modify video content then enhance motion-control-pipeline.md # Character animation end-to-end video-production.md # Full production (create + extend + audio + upscale) storyboard-to-video.md # Multi-shot storyboard → concatenated video batch-creation.md # Parallel batch generation mondo-poster-pipeline.md # End-to-end Mondo poster generation mondo-poster-to-video-pipeline.md # Animate poster into cinematic video references/ # Curated design knowledge mondo-poster/ # 37 artist styles, composition, genre templates ``` ### 功能与工作流 - **功能** 用于记录单个命令或命令组 — 包括 flag、模型、参数限制、JSON 输出格式和错误代码。 - **工作流** 将多个功能组合成端到端的 pipeline，并提供详细的分步说明。 ## 支持的模型 ### 视频模型 | 模型 | CLI 值 | 模式 | 画质 | 时长 | 宽高比 | |:---|:---|:---|:---|:---|:---| | PixVerse V6 | `v6` (默认) | Video, Transition (首/尾帧), Extend | `360p` `540p` `720p` `1080p` | `1`-`15` (任意整数) | `16:9` `4:3` `1:1` `3:4` `9:16` `3:2` `2:3` `21:9` | | PixVerse C1 | `pixverse-c1` | Video, Transition (首/尾帧), Reference | `360p` `540p` `720p` `1080p` | `1`-`15` (任意整数) | `16:9` `4:3` `1:1` `3:4` `9:16` `3:2` `2:3` | | PixVerse v5.6 | `v5.6` | Video, Transition, Reference, Extend, Motion Control | `360p` `540p` `720p` `1080p` | `1`-`10` (任意整数) | `16:9` `4:3` `1:1` `3:4` `9:16` `3:2` `2:3` | | Sora 2 | `sora-2` | Video | `720p` | `4` `8` `12` | `16:9` `9:16` | | Sora 2 Pro | `sora-2-pro` | Video | `720p` `1080p` | `4` `8` `12` | `16:9` `9:16` | | Veo 3.1 Standard | `veo-3.1-standard` | Video, Transition | `720p` `1080p` | `4` `6` `8` | `16:9` `9:16` | | Veo 3.1 Fast | `veo-3.1-fast` | Video, Transition | `720p` `1080p` | `4` `6` `8` | `16:9` `9:16` | | Veo 3.1 Lite | `veo-3.1-lite` | Video | `720p` `1080p` | `4` `5` `6` | `16:9` `9:16` | | Grok Imagine | `grok-imagine` | Video, Extend, Reference | `480p` `720p` | `1`-`15` | `16:9` `4:3` `1:1` `9:16` `3:4` `3:2` `2:3` | | Happy Horse 1.0 | `happyhorse-1.0` | Video | `720p` `1080p` | `3`-`15` | `16:9` `9:16` `1:1` `4:3` `3:4` | | Seedance 2.0 Standard | `seedance-2.0-standard` | Video, Reference, Transition | `480p` `720p` | `4`-`15` | `16:9` `4:3` `1:1` `3:4` `9:16` `21:9` | | Seedance 2.0 Fast | `seedance-2.0-fast` | Video, Reference, Transition | `480p` `720p` | `4`-`15` | `16:9` `4:3` `1:1` `3:4` `9:16` `21:9` | | Kling O3 Pro | `kling-o3-pro` | Video, Reference, Transition | `720p` | `3`-`15` | `16:9` `9:16` `1:1` | | Kling O3 Standard | `kling-o3-standard` | Video, Reference, Transition | `720p` | `3`-`15` | `16:9` `9:16` `1:1` | | Kling 3.0 Pro | `kling-3.0-pro` | Video, Transition | `720p` | `3`-`15` | `16:9` `9:16` `1:1` | | Kling 3.0 Standard | `kling-3.0-standard` | Video, Transition | `720p` | `3`-`15` | `16:9` `9:16` `1:1` | ### 图像模型 | 模型 | CLI 值 | 分辨率 | 宽高比 | |:---|:---|:---|:---| | Qwen Image | `qwen-image` | `720p` `1080p` | `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | GPT Image 2 | `gpt-image-2.0` | `1080p` `1440p` `2160p` | 取决于画质 — `1080p`: `1:1` `3:2` `2:3` · `1440p`: `1:1` `16:9` `9:16` · `2160p`: `16:9` `9:16`。需要 `--detail-level`。 | | Seedream 5.0 Lite | `seedream-5.0-lite` | `1440p` `1800p` `2160p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Seedream 4.5 | `seedream-4.5` | `1440p` `2160p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Seedream 4.0 | `seedream-4.0` | `1080p` `1440p` `2160p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Gemini 2.5 Flash (aka Nanobanana) | `gemini-2.5-flash` | `1080p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Gemini 3.0 (aka Nano Banana Pro) | `gemini-3.0` | `1080p` `1440p` `2160p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Gemini 3.1 Flash (aka Nano Banana 2) | `gemini-3.1-flash` | `512p` `1080p` `1440p` `2160p` | `auto` `1:1` `16:9` `9:16` `4:3` `3:4` `5:4` `4:5` `3:2` `2:3` `21:9` | | Kling Image O3 | `kling-image-o3` | `1080p` `1440p` `2160p` | `16:9` `9:16` `1:1` `4:3` `3:4` `3:2` `2:3` `21:9` | | Kling Image V3 | `kling-image-v3` | `1080p` `1440p` | `16:9` `9:16` `1:1` `4:3` `3:4` `3:2` `2:3` `21:9` | ## 面向 AI Agent 开发者 这些技能专为加载到 Agent 的上下文中而设计。每个技能文件都是独立的，包含： - **决策树** — 帮助 Agent 选择正确的命令 - **Flag 表格** — 包含每个参数的允许值和默认值 - **模型参考表** — 针对特定模型的参数限制 - **JSON 输出 schema** — 用于解析的精确响应格式 - **退出代码** — 确定性的错误处理 - **示例** — 可直接复制粘贴的命令 首先加载 `skills/SKILL.md` 作为入口点，然后根据需要加载特定的功能或工作流技能。 ## 社区技能 由社区基于 PixVerse CLI 构建的项目： | 项目 | 作者 | 描述 | |:---|:---|:---| | [pixverse-character-pipeline](https://github.com/Takamasa045/pixverse-character-pipeline) | [@takamasa045](https://x.com/takamasa045) | 角色驱动的视频制作 — 一个发言人图像 + YAML 配置 → 支持多语言、多宽高比、包含唇形同步、BGM 和 Remotion 渲染的数字人播报视频 | | [pixverse-shotpack](https://github.com/Takamasa045/pixverse-shotpack) | [@takamasa045](https://x.com/takamasa045) | 创意简报 → 视频镜头 pipeline — 将 Markdown 简报或 YAML 故事板转换为井然有序、可用于剪辑的 AI 生成视频资产 | ## 致谢 感谢那些为 PixVerse 生态系统做出贡献的创作者们： - [@takamasa045](https://x.com/takamasa045) — 感谢其在 PixVerse CLI 之上构建角色 pipeline 和 shotpack 制作工具 - [@vista8](https://x.com/vista8) — 感谢其 Mondo 海报设计系统，该系统中的 prompt 工程和艺术家风格库被改编用于 `pixverse:mondo-poster-design` ## 链接 - [PixVerse 官网](https://pixverse.ai) - [npm 上的 PixVerse CLI](https://www.npmjs.com/package/pixverse) - [订阅计划](https://app.pixverse.ai/subscribe)

标签：AIGC, MITM代理, 图像生成, 工作流, 提示词工程, 文档结构分析, 暗色界面, 策略决策点, 视频生成, 防御加固