Imported from jeremylongshore/tons-of-skills-marketplace (plugins/saas-packs/klingai-pack/skills/klingai-text-to-video/SKILL.md). Install upstream with npx skills add jeremylongshore/tons-of-skills-marketplace --skill klingai-text-to-video. Copyright stays with the author (MIT).

Kling AI Text-to-Video

Overview

Generate videos from text prompts using the /v1/videos/text2video endpoint. Supports models v1 through v2.6, standard/professional modes, camera control, negative prompts, and native audio (v2.6+).

Endpoint: POST https://api.klingai.com/v1/videos/text2video

Request Parameters

Complete Example — 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"}

# Create text-to-video task
response = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
    "model_name": "kling-v2-6",
    "prompt": "Aerial drone shot of a coral reef at golden hour, "
              "tropical fish swimming through crystal clear water, "
              "sun rays penetrating the surface, cinematic 4K",
    "negative_prompt": "blurry, low quality, distorted, watermark",
    "duration": "5",
    "aspect_ratio": "16:9",
    "mode": "professional",
    "cfg_scale": 0.5,
})

task = response.json()
task_id = task["data"]["task_id"]

# Poll for completion
while True:
    time.sleep(15)
    result = requests.get(
        f"{BASE}/videos/text2video/{task_id}", headers=get_headers()
    ).json()

    status = result["data"]["task_status"]
    if status == "succeed":
        video = result["data"]["task_result"]["videos"][0]
        print(f"Video URL: {video['url']}")
        print(f"Duration: {video['duration']}s")
        break
    elif status == "failed":
        raise RuntimeError(result["data"]["task_status_msg"])
    # else: submitted/processing — keep polling

With Camera Control

# Camera movement types: pan, tilt, zoom, roll
response = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
    "model_name": "kling-v2-6",
    "prompt": "A medieval castle on a cliff at sunrise, fog in the valley",
    "duration": "5",
    "mode": "standard",
    "camera_control": {
        "type": "simple",
        "config": {
            "horizontal": 5,    # pan right (negative = left), range -10 to 10
            "vertical": 0,      # tilt (negative = down, positive = up)
            "zoom": 3,          # zoom in (positive) or out (negative)
            "roll": 0,          # rotation
            "pan": 0,           # dolly left/right
            "tilt": -2,         # dolly up/down
        }
    },
})

Rule: Only one non-zero field in config for type: "simple".

With Native Audio (v2.6 only)

response = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
    "model_name": "kling-v2-6",
    "prompt": "A jazz band performing in a dimly lit club, saxophone solo, "
              "audience clapping, warm amber lighting",
    "duration": "10",
    "mode": "professional",
    "motion_has_audio": True,  # generates synchronized audio
})

Prompt Engineering Tips

Cost Reference

Error Handling

Prerequisites

• An approved brief, rights-cleared or synthetic reference material, an authorized workspace and budget, a content-policy review, and a named owner for publication and removal.

Instructions

1. Create a sandbox draft from an approved brief; do not include private individuals, protected material, or unverified claims in prompts or uploads.
2. Verify the requested duration, style, destination, credit budget, content-policy status, and draft-only setting before submission.
3. Review one watermarked canary render for policy, rights, and quality; halt on a policy or attribution concern and delete the draft rather than publishing it.
4. Promote only after owner approval and retain a redacted production receipt; remove temporary assets at the approved retention boundary.

Output

Produce a render receipt with brief ID, approved source classification, model/mode, duration, credit estimate, policy and rights-review outcome, draft destination, approver, retention/removal reference, and task ID. Exclude prompt text, identities, and credentials.

Examples

brief=synthetic-product-demo; source=rights-cleared; mode=standard; duration=5s; policy=pass; destination=draft-only; approval=pending; cleanup=24h is a safe canary request.

Resources

• Text-to-Video API
• Camera Control Guide
• Content Guidelines