Klingai Style TransferSAFE

Model-agnostic agent-skills platform with a harness-free canonical layer, verified adapters, and the ccpi package manager. Explore at tonsofskills.com.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: klingai-style-transfer
description: 'Apply artistic styles and visual effects to Kling AI video generation.
  Use when creating

  stylized content or using effects API. Trigger with phrases like ''klingai style'',
  ''kling ai effects'',

  ''klingai artistic video'', ''stylize klingai video''.

  '
allowed-tools: Read, Write, Edit, Bash(npm:*), Grep
version: 1.18.0
license: MIT
author: Jeremy Longshore <[email protected]>
tags:
- saas
- kling-ai
- style-transfer
- effects
compatibility: Designed for Claude Code
---
# Kling AI Style Transfer & Effects

## Overview

Apply artistic styles through prompt engineering, use the Effects API for pre-built visual transformations, and leverage Kolors for image-based style references. Available on v1.6+ models.

## Style via Prompt Engineering

The most direct approach -- include style descriptors in your prompt:

```python
import jwt, time, os, requests

BASE = "https://api.klingai.com/v1"

def get_headers():
    ak, sk = os.environ["KLING_ACCESS_KEY"], os.environ["KLING_SECRET_KEY"]
    token = jwt.encode(
        {"iss": ak, "exp": int(time.time()) + 1800, "nbf": int(time.time()) - 5},
        sk, algorithm="HS256", headers={"alg": "HS256", "typ": "JWT"}
    )
    return {"Authorization": f"Bearer {token}", "Content-Type": "application/json"}

# Style: Studio Ghibli watercolor
response = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
    "model_name": "kling-v2-6",
    "prompt": "A cozy cottage in a meadow, hand-painted watercolor style, "
              "soft pastel colors, Studio Ghibli aesthetic, gentle breeze",
    "negative_prompt": "photorealistic, harsh lighting, dark, gritty",
    "duration": "5",
    "mode": "professional",
    "cfg_scale": 0.7,  # higher = stricter prompt adherence
})
```

## Style Prompt Recipes

| Style | Prompt Keywords | cfg_scale |
|-------|----------------|-----------|
| Cinematic | "cinematic lighting, anamorphic lens, film grain, 35mm" | 0.5-0.6 |
| Anime | "anime style, cel-shaded, vibrant colors, clean lines" | 0.6-0.7 |
| Watercolor | "watercolor painting, soft edges, pastel, hand-painted" | 0.7-0.8 |
| Oil painting | "oil painting, thick brushstrokes, impasto, canvas texture" | 0.7-0.8 |
| Neon/cyberpunk | "neon lights, cyberpunk, rain, dark city, purple and blue" | 0.5-0.6 |
| Vintage film | "vintage 8mm film, warm tones, light leaks, soft focus" | 0.6-0.7 |
| Pixel art | "pixel art style, retro 16-bit, limited palette" | 0.8-0.9 |
| Photorealistic | "photorealistic, 4K, natural lighting, DSLR quality" | 0.4-0.5 |

## Effects API

The Effects API applies pre-built transformations to existing images. Available on v1.6+.

**Endpoint:** `POST https://api.klingai.com/v1/videos/effects`

```python
# Apply an effect to an image
response = requests.post(f"{BASE}/videos/effects", headers=get_headers(), json={
    "model_name": "kling-v1-6",
    "image": "https://example.com/portrait.jpg",
    "effect_type": "hug",           # effect to apply
    "duration": "5",
    "mode": "standard",
})

task_id = response.json()["data"]["task_id"]
# Poll for result as usual
```

## Available Effects

| Effect | Description |
|--------|-------------|
| `hug` | Embrace/hug motion between subjects |
| `kiss` | Kiss animation between subjects |
| `heart` | Heart gesture or heart-shaped framing |
| `expand` | Zoom/expand outward effect |
| `squish` | Compression/squish animation |

## Kolors Image Restyle

Use Kolors to restyle images before converting to video:

```python
# Generate styled image with Kolors
image_response = requests.post(f"{BASE}/images/kolors", headers=get_headers(), json={
    "prompt": "A cyberpunk city street, neon signs, rain-slicked roads",
    "aspect_ratio": "16:9",
    "imageCount": 1,
})

# Then use the generated image as input for I2V
image_url = image_response.json()["data"]["images"][0]["url"]
video_response = requests.post(f"{BASE}/videos/image2video", headers=get_headers(), json={
    "model_name": "kling-v2-1",
    "image": image_url,
    "prompt": "Camera slowly pushes forward through the rain, neon reflections",
    "duration": "5",
    "mode": "professional",
})
```

## cfg_scale Tuning

The `cfg_scale` parameter (0.0-1.0) controls how strictly the model follows your prompt:

| cfg_scale | Effect |
|-----------|--------|
| 0.0-0.3 | More creative freedom, may drift from prompt |
| 0.4-0.5 | Balanced (default), natural results |
| 0.6-0.7 | Stronger prompt adherence |
| 0.8-1.0 | Very strict, may reduce quality/naturalness |

**For style transfer:** Use 0.6-0.8 to ensure the style keywords are respected.

## Style Consistency Across Clips

```python
# Use a consistent style template for all clips in a project
STYLE_TEMPLATE = {
    "suffix": ", cinematic lighting, 35mm film grain, warm color grading, "
              "anamorphic lens flare, shallow depth of field",
    "negative": "cartoon, anime, painting, illustration, CGI, digital art",
    "cfg_scale": 0.6,
    "model": "kling-v2-6",
    "mode": "professional",
}

def styled_generation(scene_prompt: str):
    return requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
        "model_name": STYLE_TEMPLATE["model"],
        "prompt": scene_prompt + STYLE_TEMPLATE["suffix"],
        "negative_prompt": STYLE_TEMPLATE["negative"],
        "cfg_scale": STYLE_TEMPLATE["cfg_scale"],
        "duration": "5",
        "mode": STYLE_TEMPLATE["mode"],
    })
```

## Prerequisites

- An approved style brief, rights-cleared or synthetic references, sandbox workspace, content-policy review, credit cap, draft-only destination, and an asset-removal owner.

## Instructions

1. Use only approved style references and a watermarked sandbox canary; reject living-artist imitation requests, private materials, and unlicensed inputs unless documented rights permit them.
2. Verify reference rights, policy outcome, model/mode, duration, credit estimate, and draft destination before submission.
3. Review the ca

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.