kling-ai_skill

Overview

This skill integrates with the Kling AI API to generate high-quality images and short videos from text prompts or animate existing images. It supports text-to-image, text-to-video, and image-to-video workflows with configurable aspects like aspect ratio, duration, camera motion, and fidelity. Use it to quickly prototype visual assets or produce final media up to 1080p at 30fps.

How this skill works

Authenticate using an Access Key (AK) and Secret Key (SK) to generate a short-lived JWT used in all requests. Submit async generation tasks to endpoints for images, text-to-video, or image-to-video. Poll the provided task ID until the task status is succeed or failed, then download the resulting media URL. Camera control, fidelity, negative prompts, and callbacks are supported to tune outputs.

When to use it

• Generate concept art, hero images, or thumbnails from text prompts.
• Produce short marketing or social videos directly from descriptions.
• Animate portraits or static artwork into subtle motion videos.
• Batch-generate multiple images for A/B creative testing.
• Prototype cinematic camera moves with simple presets.

Best practices

• Keep prompts clear and under 500 characters; use negative_prompt to exclude unwanted artifacts.
• Start with image_fidelity ~0.5 and adjust toward 0 or 1 to trade creativity vs. faithfulness.
• Use short polling intervals (10s) and handle failed states with retries or fallbacks.
• Prefer pro mode for portrait or high-detail video outputs (1080p) and std for faster, cheaper results.
• Encode images as base64 or provide a stable URL when using image-to-video to avoid upload issues.

Example use cases

• Text-to-image: create 16:9 hero images for blog headers or ads.
• Text-to-video: generate 5–10s cinematic clips for social media from a script line.
• Image-to-video: animate a portrait with natural blink and head turns for an about page.
• Batch image generation: produce multiple style variants in parallel for creative review.
• Cinematic workflow: apply camera_presets (zoom, pan, orbit) to create dynamic scene reveals.