Metadata

name: klingai-video
description: Official Kling AI Skill. Call Kling AI video generation API; supports text-to-video, image-to-video, Omni (multi-modal), and multi-shot. Script auto-selects the best endpoint by input. Use when the user mentions "text-to-video", "image-to-video", "Kling", "multi-shot", "shots", "subject", "可灵", "文生视频", "图生视频", "多镜头", "分镜", "主体", "动画生成", "ビデオ生成", "영상 생성".
metadata: {"openclaw":{"emoji":"🎬","requires":{"bins":["node"]},"primaryEnv":"KLING_TOKEN","homepage":"https://app.klingai.com/cn/dev/document-api"}}

Language: Always respond to the user in their own language. Script CLI output is bilingual (English / Chinese). Present results in the user's language.

Kling AI Video Generation (Kling Official)

Official Kling AI skill for agents to call Kling video API. Single entry for text-to-video, image-to-video, Omni (image/video/subject), and multi-shot. The script picks the API by parameters; zero external dependencies, Node 18+.

Prerequisites

Node.js 18+
Auth (choose one): KLING_TOKEN (recommended) or KLING_API_KEY (format accessKey|secretKey)
If KLING_API_BASE is not set, the script probes China (Beijing) / Global (Singapore) and caches to ~/.config/kling/state.json
To force a region: set KLING_API_BASE
After switching account or region: rm ~/.config/kling/state.json

Use cases

Use for:

Text-to-video (text description only)
Image-to-video (first frame + optional last frame)
Omni: text + image + reference video + subject (create subject first via klingai-element)
Multi-shot video

Not for:

Image generation (text/image-to-image) → klingai-image
Create or manage subjects → klingai-element

Quick start

# Show all options
node {baseDir}/scripts/kling_video.mjs --help

# Text-to-video
node {baseDir}/scripts/kling_video.mjs \
  --prompt "A cat running on the grass, sunny day" \
  --output_dir ./output

# Image-to-video
node {baseDir}/scripts/kling_video.mjs \
  --image ./photo.jpg \
  --prompt "Wind blowing hair" \
  --output_dir ./output

# Omni (first frame + subject)
node {baseDir}/scripts/kling_video.mjs \
  --prompt "<<<element_1>>> walking in the park" \
  --image ./scene.jpg \
  --element_ids 123456

# Multi-shot
node {baseDir}/scripts/kling_video.mjs \
  --multi_shot --shot_type customize \
  --multi_prompt '[{"index":1,"prompt":"Sunrise scene","duration":"5"},{"index":2,"prompt":"Close-up","duration":"5"}]' \
  --duration 10

# Query existing task
node {baseDir}/scripts/kling_video.mjs --task_id <id> --download

Core parameters

Parameter	Description	Default
`--prompt`	Video description (required for text/image/Omni)	—
`--image`	First-frame image path or URL (comma-separated multiple → Omni)	—
`--duration`	Duration 3–15 s	5
`--model`	kling-v3 / kling-v3-omni / kling-video-o1 (script chooses)	by endpoint
`--mode`	pro (1080P) / std (720P)	pro
`--aspect_ratio`	16:9 / 9:16 / 1:1	16:9
`--sound`	on / off. Supported by kling-v3/omni; with `--video` only off; kling-video-o1 does not support sound	off
`--image_tail`	Last-frame image (image-to-video / Omni)	—
`--element_ids`	Subject IDs, comma-separated (Omni)	—
`--video`	Reference video path or URL (Omni)	—
`--multi_shot`	Enable multi-shot	false
`--shot_type`	Shot type: customize	—
`--multi_prompt`	Shots JSON array (max 6)	—
`--output_dir`	Output directory	./output

Run --help for full parameters.

API routing (script auto-selects)

Text only → /v1/videos/text2video (kling-v3)
Single first-frame image, no subject/multi-shot/reference video → /v1/videos/image2video (kling-v3)
Multiple images / subject / reference video / multi-shot → /v1/videos/omni-video (kling-v3-omni or kling-video-o1)

Prompt template syntax (Omni)

In the prompt, reference inputs with <<<>>>:

<<<image_1>>> — first image from --image
<<<element_1>>> — first subject from --element_ids
<<<video_1>>> — video from --video

Notes

Generation typically 1–5 minutes; script polls by default
In multi-shot mode --prompt is ignored; use --multi_prompt for each shot
Reference video and audio generation are mutually exclusive
API details in reference.md

Klingai Video

安装技能