kling-video-generator_skill

Overview

This skill generates and manipulates high-quality videos using the Kling 3.0 Omni model. It supports text-to-video, image-to-video, video editing, reference-guided generation, multi-shot sequences, and synchronized audio. The skill provides a clear workflow to build compliant API payloads and utilities for authentication and task polling.

How this skill works

I convert user intent into well-formed prompts and a validated JSON payload that respects Kling Omni parameter rules. The process splits into prompt design (single- or multi-shot, template references for images/videos) and parameter construction (setting fields like refer_type, shot_type, duration, sound, aspect_ratio, and image/video lists). A provided Python helper handles JWT auth, task creation, and polling to return final video URLs.

When to use it

• Create a new video from a textual script or concept (text-to-video).
• Animate a static image into a short video with descriptive motion (image-to-video).
• Edit or restyle an existing video while preserving or changing content (video-to-video editing).
• Generate a multi-shot sequence or storyboard-driven video.
• Produce videos with synchronized speech or sound effects (audio-synced output).

Best practices

• Always craft prompts following the official prompt formula and use explicit template tokens for any reference image or video (e.g., <<<image_1>>>).
• Decide single-shot vs multi-shot up front; multi-shot requires shot_type "customize" and a matching multi_prompt array.
• Explicitly set refer_type: use "base" for editing and "feature" for reference-driven generation. Do not rely on defaults.
• Ensure duration handling: duration is ignored in base (editing) mode and must equal the sum of multi_prompt durations in customize mode.

Example use cases

• Turn a short screenplay paragraph into a single-shot promotional clip with generated audio.
• Animate a product image into a 10s motion demo with camera-like movement borrowed from a reference video.
• Edit a recorded clip to change the scene’s art style and swap background audio using editing mode.
• Create a 4-shot social media sequence with distinct prompts per shot and a fixed total duration.
• Generate an explainer video where synthetic speech is aligned to on-screen actions and transitions.