Vidu
调用vwu.ai平台上的vidu系列模型,支持7种型号,兼容OpenAI API格式,需配置vwu.ai API密钥使用。
MIT-0 · Free to use, modify, and redistribute. No attribution required.
⭐ 0 · 30 · 0 current installs · 0 all-time installs
MIT-0
Security Scan
OpenClaw
Suspicious
high confidencePurpose & Capability
The SKILL.md and the included scripts implement a Vidu video-generation client that uses an API key (VWU_API_KEY) and HTTP calls to vwu.ai — this matches the described purpose. However, the registry metadata claims no required env vars or binaries even though the scripts require VWU_API_KEY and rely on common CLI tools (curl, jq, base64, date, awk, open/ffplay). The omission of these requirements in metadata is an incoherence.
Instruction Scope
The SKILL.md stays within the video-generation scope: create tasks, poll status, download content from the documented API endpoints. It instructs users to set VWU_API_KEY and provides sample commands. The scripts read only local files (models.txt, images) and call the API endpoints; they do not attempt to read unrelated system files. Minor point: the vwu-chat.sh prints the first 8 characters of the API key in error messages, which is credential exposure in logs.
Install Mechanism
There is no install spec and no remote download; this is instruction/script-only. That limits installation risk — nothing is being fetched or executed from unknown URLs by the skill itself.
Credentials
Only VWU_API_KEY is actually needed by the code, which is proportionate to the purpose, but the skill metadata does not declare this required env var. The scripts also accept VWU_BASE_URL (with a default) — allowing overriding the endpoint could redirect credentials to an arbitrary server if misconfigured. The scripts also echo a truncated API key on errors (revealing part of the secret in terminal output).
Persistence & Privilege
The skill does not request always:true and does not modify other skills or system-wide settings. Autonomous invocation is allowed by default (normal for skills) and there are no signs the skill attempts to persist itself beyond its own files.
What to consider before installing
This skill appears to be what it says (a client for vwu.ai Vidu video models) but has metadata inconsistencies you should address before trusting it: 1) The scripts require an API key (VWU_API_KEY) and CLI tools (curl, jq, base64, etc.), yet the registry metadata lists no env vars or binaries — verify these requirements locally. 2) Inspect the scripts before use; they send your API key to VWU_BASE_URL and will follow whatever URL is set there, so do not set VWU_BASE_URL to an untrusted host. 3) Be aware the scripts print the first 8 characters of your key on some errors (logs/terminal), which is minor exposure — avoid pasting real keys into shared logs. 4) Because the package includes executable shell scripts, run them in a controlled environment or review them carefully before executing. If you need higher assurance, ask the publisher to update the registry metadata to declare VWU_API_KEY and required binaries, and to remove or document any behavior that prints secrets or allows endpoint overrides.Like a lobster shell, security has layers — review code before you run it.
Current versionv1.1.0
Download zipaiimage-to-videolatestllmvideoviduvwu
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
SKILL.md
vwu-vidu Models for vwu.ai
vwu.ai 平台上的 Vidu 视频生成模型技能。
✅ 支持的模型
已验证可用:
- viduq3-pro - Vidu Q3 Pro(推荐,完全支持)
- viduq2-turbo - Vidu Q2 Turbo(需测试)
不支持:
- vidu2.0, viduq1, viduq1-classic - 旧版本已下线
- viduq2-pro - API 不支持
🎬 功能特性
文生视频
从文本描述生成视频:
vwu-video --model viduq3-pro \
--prompt "一只可爱的猫咪在跳舞" \
--duration 5 \
--wait
图生视频
从静态图片生成动态视频:
vwu-video --model viduq3-pro \
--image photo.jpg \
--prompt "让图片中的人物跳舞" \
--duration 5 \
--wait
⚙️ 配置
1. 获取 API Key
重要: 使用前必须在 vwu.ai 控制台创建 API Key
- 访问 https://vwu.ai
- 登录账号
- 进入**「令牌管理」**
- 点击**「新建令牌」**
- 设置名称(如:claw-video)
- 复制生成的密钥
2. 设置环境变量
export VWU_API_KEY="sk-your-key-here"
建议添加到 shell 配置文件 (~/.zshrc 或 ~/.bashrc):
echo 'export VWU_API_KEY="sk-your-key-here"' >> ~/.zshrc
source ~/.zshrc
📖 使用示例
基础文生视频
# 最简单的用法
vwu-video --prompt "猫咪跳舞" --wait
# 指定参数
vwu-video \
--model viduq3-pro \
--prompt "一朵花在风中摇曳,4K高清" \
--duration 5 \
--ratio 16:9 \
--output flower.mp4 \
--wait
图生视频
# 使用图片作为参考
vwu-video \
--image /path/to/photo.jpg \
--prompt "让图片中的人物开始跳舞" \
--duration 5 \
--output dance.mp4 \
--wait
查询任务状态
# 查询任务状态
vwu-video --status 931029583942135808
# 下载已完成的视频
vwu-video --download 931029583942135808 --output video.mp4
高级参数
vwu-video \
--model viduq3-pro \
--prompt "电影级风景延时" \
--duration 10 \
--ratio 16:9 \
--metadata '{"resolution":"1080p","fps":24}' \
--wait
📋 参数说明
选项
| 参数 | 说明 | 默认值 |
|---|---|---|
--model | 模型名称 | viduq3-pro |
--prompt | 视频描述(必需) | - |
--image | 参考图片路径(可选) | - |
--duration | 视频时长(秒) | 5 |
--ratio | 宽高比 | 16:9 |
--output | 输出文件名 | video_时间戳.mp4 |
--wait | 等待生成完成并下载 | false |
--status | 查询任务状态 | - |
--download | 下载已完成的视频 | - |
--list-models | 列出可用模型 | - |
宽高比选项
16:9- 横屏(默认)9:16- 竖屏(手机视频)1:1- 方形(社交媒体)
metadata 字段
{
"duration": 5, // 时长 2-10 秒
"aspect_ratio": "16:9", // 宽高比
"resolution": "1080p", // 分辨率(可选)
"fps": 24, // 帧率(可选)
"style": "realistic" // 风格(可选)
}
🎯 提示词建议
好的提示词特征
-
详细描述场景
- ✅ "一只橘色的猫咪坐在木地板上,头部左右摇摆,阳光从窗户射入"
- ❌ "猫跳舞"
-
指定动作和风格
- ✅ "人物优雅地旋转,手臂舒展,芭蕾舞风格,慢动作"
- ❌ "人跳舞"
-
添加质量关键词
- ✅ "4K高清,电影级质量,专业摄影,光线柔和"
- ❌ "视频"
示例提示词
- "一只柯基犬在草地上奔跑,尾巴摇摆,阳光明媚"
- "城市夜景,车流延时摄影,灯光流动,高质量"
- "一朵玫瑰花慢慢绽放,花瓣层层展开,微距拍摄"
- "海浪拍打礁石,水花四溅,慢动作,日落时分"
⚠️ 限制和注意事项
API Key 额度
- 每次生成会消耗 API key 额度
- 如提示"额度不足",请在 vwu.ai 控制台充值
- 或创建新的 API 令牌
生成时间
- 通常需要 2-5 分钟
- 取决于服务器负载和视频时长
- 使用
--wait参数会自动等待完成
视频参数
- 时长: 建议使用 5 秒,最长 10 秒
- 分辨率: 默认 1080p,可达 4K
- 格式: 输出为 MP4 格式
🔄 API 信息
端点: https://api.vwu.ai/v1/videos 文档: https://platform.vidu.cn/docs/image-to-video 模式: 异步任务(创建 → 查询 → 下载)
任务状态
queued- 队列中in_progress- 生成中succeeded/completed- 完成failed- 失败
💡 故障排除
常见问题
Q: 提示"未设置 VWU_API_KEY"
A: 需要先创建 API Key
1. 访问 https://vwu.ai
2. 登录 → 令牌管理 → 新建令牌
3. 设置环境变量: export VWU_API_KEY='your-key'
Q: 生成失败,返回错误
A: 检查以下几点:
1. API key 额度是否充足
2. 提示词是否合规
3. 图片格式是否正确(JPG/PNG)
4. 查看详细错误信息
Q: 视频生成时间过长
A: 正常情况:
- 文生视频: 2-4 分钟
- 图生视频: 3-5 分钟
- 如超过 10 分钟未完成,可能失败
Q: 生成的视频质量不满意
A: 优化提示词:
1. 更详细地描述场景
2. 添加质量关键词(4K、电影级等)
3. 指定风格和动作
4. 尝试不同的宽高比
📞 相关链接
- vwu.ai 控制台: https://vwu.ai
- Vidu 官方文档: https://platform.vidu.cn/docs/image-to-video
- New API 文档: https://docs.newapi.ai/en
- ClawHub: https://clawhub.com
🆕 更新日志
v1.1.0 (2026-03-16)
- ✅ 修正 API 端点(api.vwu.ai)
- ✅ 添加图生视频支持
- ✅ 添加完整的 vwu-video.sh 脚本
- ✅ 更新模型列表(移除不支持的模型)
- ✅ 添加 API Key 配置指引
v1.0.0 (2026-03-16)
- 初始版本
技能维护: Claw AI Assistant 最后更新: 2026-03-16 版本: 1.1.0
Files
4 totalSelect a file
Select a file to preview.
Comments
Loading comments…
