技能详情(站内镜像,无评论)
作者:徐京徽 @a574824551
许可证:MIT-0
MIT-0 ·免费使用、修改和重新分发。无需归因。
版本:v1.0.0
统计:⭐ 2 · 71 · 0 current installs · 0 all-time installs
⭐ 2
安装量(当前) 0
🛡 VirusTotal :良性 · OpenClaw :良性
Package:a574824551/polecomic
安全扫描(ClawHub)
- VirusTotal :良性
- OpenClaw :良性
OpenClaw 评估
The skill is internally consistent with its stated purpose (generating storyboard images and videos via the Volcengine/’Ark’ API) but has a few minor metadata and documentation mismatches you should be aware of before running it.
目的
The skill's name, description, SKILL.md, and included scripts all align: they implement a multi-stage manga/animation pipeline (script → storyboard → character prompts → image generation → video generation) using the Volcengine/Ark APIs. One mismatch: the skill metadata in the registry claims no required environment variables, but the runtime docs and both scripts clearly require ARK_API_KEY for API authentication.
说明范围
SKILL.md gives explicit stage-by-stage instructions and mandates pausing for user confirmation between stages. The instructions direct the agent and the operator to run the provided scripts, update CHARACTER_PROMPTS in call_image_api.py, and populate storyboard.json; they do not instruct reading unrelated system files or exfiltrating secrets. The scripts read/write only project files (storyboard.json, output/*, generation_log.json) and call th…
安装机制
There is no automated install spec; the skill is instruction + Python scripts and requires the public Python packages 'volcengine-python-sdk[ark]' and 'requests' (installed via pip). No downloads from unknown URLs or archive extraction are present. This is a low-risk, typical requirement for a scripts-based skill.
证书
Only one credential is used at runtime: ARK_API_KEY (the scripts look for it in the environment or as a hardcoded variable). That is proportionate to the stated purpose. However, the registry metadata did not list required env vars while the documentation and scripts require ARK_API_KEY — this mismatch should be resolved before install so users know an external API key is mandatory.
持久
The skill does not request 'always: true' and does not modify other skills or global agent configuration. It performs normal file I/O in its own workspace (output/, storyboard.json) and requires manual edits to CHARACTER_PROMPTS for character images; nothing in the package implies elevated or hidden persistence.
综合结论
This skill appears to implement exactly what it claims: a multi-stage manga/animation pipeline that calls Volcengine/‘Ark’ image and video APIs. Before installing/running: 1) Be aware you must supply an ARK_API_KEY (the registry metadata omitted this); the scripts will exit if it's not set. 2) The scripts will send your prompts, images (embedded as base64), and storyboard data to the Ark service (https://ark.cn-beijing.volces.com/api/v3) — onl…
安装(复制给龙虾 AI)
将下方整段复制到龙虾中文库对话中,由龙虾按 SKILL.md 完成安装。
请把本段交给龙虾中文库(龙虾 AI)执行:为本机安装 OpenClaw 技能「manga-animation」。简介:完整的漫剧(动画漫画)生成工作流 skill。当用户想要:创作漫画剧集、生成动画短片、制作分镜脚本、设计原创角色、生成动漫风格视频片段、将故事想法转化为可视化…。
请 fetch 以下地址读取 SKILL.md 并按文档完成安装:https://raw.githubusercontent.com/openclaw/skills/refs/heads/main/skills/a574824551/polecomic/SKILL.md
(来源:yingzhi8.cn 技能库)
SKILL.md
---
name: manga-animation
description: >
完整的漫剧(动画漫画)生成工作流 skill。当用户想要:创作漫画剧集、生成动画短片、制作分镜脚本、设计原创角色、生成动漫风格视频片段、将故事想法转化为可视化内容,或任何涉及"漫剧""动漫创作""分镜""人设""画风设定""AI生成动画""参考图生成"的请求时,必须使用此 skill。
流程覆盖:剧本创作 → 分镜设计 → 人设+画风设定 → 分镜图生成(方舟图像)→ 视频生成(方舟图生视频)。每个阶段结束后必须等待用户确认再继续。即使用户只说"给我做个漫画动画"也应立即触发本 skill。
compatibility: "需要 bash_tool、create_file。Python 依赖:pip install 'volcengine-python-sdk[ark]' requests"
---
# 漫剧生成 Skill
## 核心原则
**每个阶段完成后必须暂停,展示产物,等用户确认或修改后再进入下一阶段。**
禁止一口气跑完所有阶段——用户需要在每个关卡看到结果并决定是否调整。
```
[阶段1] 剧本 ──→ ⏸ 用户确认
↓
[阶段2] 分镜表 ──→ ⏸ 用户确认
↓
[阶段3] 人设 + 画风 ──→ ⏸ 用户确认
↓
[阶段4] 分镜图生成(call_image_api.py)──→ ⏸ 用户确认图像质量
↓
[阶段5] 视频生成(call_video_api.py)
```
---
## 技术栈
| 用途 | 模型 | SDK 调用方式 |
|------|------|------------|
| 图像(人设图 + 分镜图) | `doubao-seedream-5-0-260128` | `client.images.generate` |
| 视频(图生视频) | `doubao-seedance-1-0-lite-i2v-250428`(调试)<br>`doubao-seedance-1-0-pro-250528`(正式) | `client.content_generation.tasks.create` |
**只需一个 Key**:`ARK_API_KEY`(方舟平台统一鉴权)
```bash
export ARK_API_KEY="your_key_here"
# 或在脚本顶部直接填写 ARK_API_KEY = "..."
```
---
## 阶段1:剧本创作
**参考:** `references/script-template.md`
- 若用户提供**完整剧本** → 跳过,直接进阶段2
- 若用户提供**创意/想法** → 输出:世界观、角色清单、三幕大纲、分场景剧本(含台词/动作/情绪)
保存:`output/[剧名]/script.md`
**⏸ 检查点**:展示剧本,询问是否调整,确认后进阶段2。
---
## 阶段2:分镜设计
**参考:** `references/storyboard-guide.md`
同时输出两个文件:
### storyboard.md(供人阅读)
完整分镜表格,见 storyboard-guide.md
### storyboard.json(供脚本调用)
```json
{
"title": "剧名",
"art_style_prefix": "",
"shots": [
{
"shot_id": "S01_01",
"duration": 4,
"description_cn": "中文画面说明",
"image_prompt": "静态画面的英文描述,不含运动",
"video_prompt": "运动方式描述,如 character walking slowly forward",
"characters": ["CHAR_001"],
"dialogue": "台词/音效",
"reference_image": null
}
]
}
```
**image_prompt vs video_prompt 分工:**
- `image_prompt`:描述画面构成(景别、角色、场景、光线),用于生成静态分镜图
- `video_prompt`:只描述运动(镜头运动 + 角色动作),图生视频时用
**⏸ 检查点**:展示分镜表,提示用户此处调整最省钱,确认后进阶段3。
---
## 阶段3:人设 + 画风设定
**参考:** `references/character-guide.md`
### 3A. 输出角色人设卡:`characters/CHAR_001.md`
人设卡必须包含可直接用于 Prompt 的精确外貌描述,以及完整的图像生成 Prompt。示例结构:
```markdown
## CHAR_001 — [角色名]
年龄/性别/身份/性格关键词
### 外貌
发型发色 / 眼睛 / 体型 / 标志性特征(疤痕/纹路/配件)
### 服装(日常 / 战斗)
### 图像生成 Prompt(英文,完整可直接使用)
17-year-old teenage boy, [detailed appearance], [clothing],
full body front view, white background,
anime style, cel shading, shonen manga, high quality, masterpiece, sharp focus
ref_image_path:
```
### 3B. 确定画风,填入 storyboard.json
```json
"art_style_prefix": "anime style, shonen manga, cel shading, vibrant colors, high quality, masterpiece"
```
### 3C. ⚡ 关键步骤:更新脚本的 CHARACTER_PROMPTS
将角色 Prompt 填入 `call_image_api.py` 顶部的字典,每次有新角色都要更新:
```python
CHARACTER_PROMPTS = {
"CHAR_001": "17-year-old teenage boy, black short hair, ..., full body front view, white background, anime style, ...",
"CHAR_002": "...",
}
```
**⏸ 检查点**:展示所有人设卡 + 画风设定,确认 Prompt 准确描述了目标形象后进阶段4(开始消耗 API 额度)。
---
## 阶段4:分镜图生成
**脚本:** `scripts/call_image_api.py`
**API:** 方舟 `doubao-seedream-5-0-260128`,size=`2K`
### 运行命令
```bash
# Step 1:生成角色人设参考图
python call_image_api.py characters
--output output/[剧名]/characters/images/
# Step 2:生成分镜图(读取 storyboard.json 的 art_style_prefix + image_prompt)
python call_image_api.py storyboard
--input output/[剧名]/storyboard.json
--output output/[剧名]/storyboard/frames/
```
脚本完成后自动将帧图路径写回 `storyboard.json` 的 `reference_image` 字段。
**⏸ 检查点**:引导用户打开 `storyboard/frames/` 查看图像,重点检查:
1. 角色外貌是否符合人设
2. 场景氛围/构图是否正确
3. 记录需要重生成的 shot_id(可单独重跑)
确认后进阶段5。
---
## 阶段5:视频生成
**脚本:** `scripts/call_video_api.py`
**API:** 方舟 `doubao-seedance-1-0-lite-i2v-250428`(lite,调试)
前置条件:`storyboard.json` 中所有 shot 的 `reference_image` 已填写。
### 运行命令
```bash
python call_video_api.py batch
--input output/[剧名]/storyboard.json
--output output/[剧名]/videos/
# 如图片目录非默认,加 --frames-dir
python call_video_api.py batch
--input output/[剧名]/storyboard.json
--output output/[剧名]/videos/
--frames-dir output/[剧名]/storyboard/frames/
```
脚本工作流:提交任务 → 每8s轮询 → 成功后下载 .mp4 → 更新 generation_log.json。
### lite vs pro 切换
```python
# call_video_api.py 顶部修改这一行
ARK_VIDEO_MODEL = "doubao-seedance-1-0-lite-i2v-250428" # 调试
ARK_VIDEO_MODEL = "doubao-seedance-1-0-pro-250528" # 正式出片
```
---
## 输出文件结构
```
output/[剧名]/
├── script.md
├── storyboard.md
├── storyboard.json ← 核心,贯穿阶段2-5
├── art_style.md
├── characters/
│ ├── CHAR_001.md
│ └── images/
│ └── CHAR_001_front.png
├── storyboard/
│ └── frames/
│ ├── S01_01.png ← 阶段4产物 = 阶段5首帧
│ └── ...
├── videos/
│ └── S01_01.mp4
└── generation_log.json
```
---
## 常见问题快查
| 问题 | 解决方式 |
|------|---------|
| `❌ 未配置 ARK_API_KEY` | 脚本顶部填写,或 `export ARK_API_KEY=xxx` |
| `❌ 找不到参考图` | 先完成阶段4生成分镜图;或用 `--frames-dir` 手动指定 |
| 视频人物外貌与分镜图不符 | `video_prompt` 只描述动作,不重复角色外貌 |
| 图像风格不统一 | 检查 `storyboard.json` 的 `art_style_prefix` 是否已填写 |
| `volcengine SDK` 未安装 | `pip install 'volcengine-python-sdk[ark]' requests` |