Kling Video Generation
GitHub用于通过 Kling API 生成 AI 视频、将图片转为视频或查询任务状态。支持文本转视频和图片转视频，自动处理任务等待与结果获取，默认使用 kling-v3 模型及 pro 模式。
Trigger Scenarios
Install
npx skills add fanfan-de/anybox --skill Kling Video Generation -g -y
SKILL.md
Frontmatter
{
"name": "Kling Video Generation",
"description": "Use when the user wants to generate, inspect, or wait for AI video generation tasks through the Kling API."
}
Kling Video Generation
Use the Kling MCP tools when the user asks to generate AI videos with Kling, turn an image into a video, or check a Kling video generation task.
Workflow
- For a pure prompt, call
kling_create_text_to_video. - For an image prompt, call
kling_create_image_to_video. - Return the
task_idimmediately if the user only asked to start generation. - If the user expects the final video, call
kling_wait_for_videowith the same endpoint and task ID. - When the task succeeds, show the returned video URL(s) from
task.urls.
Defaults
Use these defaults unless the user asks otherwise:
model_name:kling-v3mode:produration:5aspect_ratio:16:9
Only set sound, cfg_scale, or model-specific controls when the user asks for them or when they are clearly needed. Some Kling models do not support every parameter.
Images
For image-to-video, image and image_tail must be public URLs or bare base64 image content. Do not pass a data:image/...;base64, prefix.
Results
Kling generation URLs expire after the provider's retention window. If the user needs a durable artifact, download or transfer the returned video URL using available file or storage tools after the task succeeds.
Version History
- 08dc189 Current 2026-07-05 19:02
