Kling AI Storage Integration
Overview
Kling AI video URLs from task_result.videos[].url are temporary CDN links that expire. You must download and store videos in your own storage. This skill covers S3, GCS, and Azure Blob.
Download from Kling CDN
import requests
import os
def download_video(video_url: str, output_dir: str = "output") -> str:
"""Download generated video from Kling CDN."""
os.makedirs(output_dir, exist_ok=True)
# Extract filename or generate one
filename = video_url.split("/")[-1].split("?")[0]
if not filename.endswith(".mp4"):
filename = f"kling_{int(time.time())}.mp4"
filepath = os.path.join(output_dir, filename)
response = requests.get(video_url, stream=True, timeout=120)
response.raise_for_status()
with open(filepath, "wb") as f:
for chunk in response.iter_content(chunk_size=8192):
f.write(chunk)
size_mb = os.path.getsize(filepath) / (1024 * 1024)
print(f"Downloaded: {filepath} ({size_mb:.1f} MB)")
return filepath
Upload to AWS S3
import boto3
def upload_to_s3(filepath: str, bucket: str, key_prefix: str = "kling-videos/") -> str:
"""Upload video to S3 and return public URL."""
s3 = boto3.client("s3")
filename = os.path.basename(filepath)
s3_key = f"{key_prefix}{filename}"
s3.upload_file(
filepath, bucket, s3_key,
ExtraArgs={"ContentType": "video/mp4", "CacheControl": "max-age=86400"}
)
url = f"https://{bucket}.s3.amazonaws.com/{s3_key}"
print(f"Uploaded to S3: {url}")
return url
# Generate signed URL for private buckets
def get_signed_url(bucket: str, key: str, expiry: int = 3600) -> str:
s3 = boto3.client("s3")
return s3.generate_presigned_url(
"get_object",
Params={"Bucket": bucket, "Key": key},
ExpiresIn=expiry,
)
Upload to Google Cloud Storage
from google.cloud import storage
def upload_to_gcs(filepath: str, bucket_name: str, prefix: str = "kling-videos/") -> str:
"""Upload video to GCS and return public URL."""
client = storage.Client()
bucket = client.bucket(bucket_name)
filename = os.path.basename(filepath)
blob = bucket.blob(f"{prefix}{filename}")
blob.upload_from_filename(filepath, content_type="video/mp4")
blob.make_public() # or use signed URLs for private access
print(f"Uploaded to GCS: {blob.public_url}")
return blob.public_url
# Signed URL for private access
def get_gcs_signed_url(bucket_name: str, blob_name: str, expiry_min: int = 60) -> str:
from datetime import timedelta
client = storage.Client()
bucket = client.bucket(bucket_name)
blob = bucket.blob(blob_name)
return blob.generate_signed_url(expiration=timedelta(minutes=expiry_min))
Upload to Azure Blob Storage
from azure.storage.blob import BlobServiceClient
def upload_to_azure(filepath: str, container: str,
connection_string: str = None) -> str:
"""Upload video to Azure Blob Storage."""
conn_str = connection_string or os.environ["AZURE_STORAGE_CONNECTION_STRING"]
client = BlobServiceClient.from_connection_string(conn_str)
filename = os.path.basename(filepath)
blob_client = client.get_blob_client(container=container, blob=f"kling-videos/{filename}")
with open(filepath, "rb") as f:
blob_client.upload_blob(f, content_type="video/mp4", overwrite=True)
url = blob_client.url
print(f"Uploaded to Azure: {url}")
return url
End-to-End Pipeline
def generate_and_store(prompt: str, bucket: str, provider: str = "s3"):
"""Generate video with Kling AI and store in cloud."""
# 1. Generate
r = requests.post(f"{BASE}/videos/text2video", headers=get_headers(), json={
"model_name": "kling-v2-master",
"prompt": prompt,
"duration": "5",
"mode": "standard",
}).json()
task_id = r["data"]["task_id"]
# 2. Poll
result = poll_task("/videos/text2video", task_id)
video_url = result["videos"][0]["url"]
# 3. Download
filepath = download_video(video_url)
# 4. Upload
if provider == "s3":
return upload_to_s3(filepath, bucket)
elif provider == "gcs":
return upload_to_gcs(filepath, bucket)
elif provider == "azure":
return upload_to_azure(filepath, bucket)
# 5. Cleanup temp file
os.remove(filepath)
Metadata Preservation
import json
def save_with_metadata(filepath: str, task_id: str, prompt: str, model: str):
"""Save video metadata alongside the file."""
meta = {
"task_id": task_id,
"prompt": prompt,
"model": model,
"generated_at": time.strftime("%Y-%m-%dT%H:%M:%SZ"),
"filename": os.path.basename(filepath),
}
meta_path = filepath.replace(".mp4", ".meta.json")
with open(meta_path, "w") as f:
json.dump(meta, f, indent=2)
return meta_path
Prerequisites
- An approved storage destination, encryption and retention policy, rights-cleared or synthetic draft asset, least-privilege service identity, metadata-redaction rules, and a tested deletion path.
Instructions
- Upload only watermarked draft canaries to an allowlisted sandbox bucket; reject public ACLs, unapproved regions, or assets without rights and policy clearance.
- Encrypt at rest and in transit, store only redacted metadata, and verify destination, access scope, retention, and removal controls before approval.
- Halt uploads on permission, policy, rights, region, or retention drift; delete staged assets and restore the prior storage configuration.
- Promote only after owner approval and preserve a redacted receipt rather than prompt text, asset URLs, or identifying metadata.
Output
Produce a storage receipt with asset classification, destination/region classification, encryption and access checks, policy/rights outcome, draft-only status, retention/deletion proof, approver, and rollback reference. Exclude prompts, URLs, identities, and credentials.
Error Handling
| Condition | Response |
|---|---|
| Destination, ACL, or region is unapproved | Stop the transfer, delete staged copies, and restore the approved destination policy. |
| Rights, policy, or retention control fails | Quarantine and remove the draft; require owner review before resubmission. |
Examples
asset=synthetic-draft; destination=approved-sandbox; encryption=pass; acl=private; policy=pass; retention=24h; deletion=tested supports a controlled upload.