Reference to Video

Kling 4.0 puts the people, objects or scenes in your reference images into a newly generated video. You can also add reference videos, for example for the subject to copy the moves in one.

Create a task

Use the shared create endpoint with model: "kling-4.0". It usually responds in under 1 second and the cost is frozen within a few seconds; the video usually takes 1–5 minutes. Get the result once the task has finished.

Headers

Body

Send at least one of image_urls and video_urls (reference videos alone are fine); with neither, the request returns 400 invalid_request with param set to image_urls. video_url isn't accepted: sending it returns 400 invalid_request with param set to video_url.

Two reference images, portrait 1080p, 5 seconds:

curl https://api.nanoapi.ai/v1/tasks \
  -H "Authorization: Bearer $NANOAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-4.0",
    "mode": "reference_to_video",
    "image_urls": [
      "https://example.com/cat.jpg",
      "https://example.com/car.jpg"
    ],
    "prompt": "The cat sits on the hood of the car",
    "aspect_ratio": "9:16",
    "resolution": "1080p",
    "duration": 5
  }'

A reference image and a reference video (the cat walks like the woman in the video), 16:9 720p, 5 seconds:

curl https://api.nanoapi.ai/v1/tasks \
  -H "Authorization: Bearer $NANOAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-4.0",
    "mode": "reference_to_video",
    "image_urls": ["https://example.com/cat.jpg"],
    "video_urls": ["https://example.com/walk.mp4"],
    "prompt": "The cat from @Image1 walks down the street like the woman in @Video1",
    "aspect_ratio": "16:9",
    "resolution": "720p",
    "duration": 5
  }'

Response: 200 OK. data is the newly created task, with only these fields; each one is described in Create a task: Response. Keep data.id to get the result: once the task succeeds, the video is at data.outputs[0].url of the retrieved task or the webhook. The task can still fail after it's created (for example a file can't be downloaded), so go by what retrieving the task or the webhook tells you.

{
  "code": 200,
  "message": "success",
  "data": {
    "object": "task",
    "id": "tk_01jbc4k7m2n9p5q8r3s6t1v4w7",
    "model": "kling-4.0",
    "status": "queued",
    "error": null,
    "created_at": "2026-10-30T08:30:00Z"
  }
}

Pricing

cost = duration × price per second of the resolution

Loading the current prices…

Example: 1080p for 5 seconds (the first request above) costs 5 × the 1080p price per second.

With reference videos (video_urls) the task is billed at the video_edit price per second instead; GET /v1/products has no separate row for it. The second request above (720p, 5 seconds) costs 5 × the video_edit 720p price per second:

Loading the current prices…

Billed seconds are the duration you ask for, however many images or videos you send and however long the videos are. Prices change, so read the current ones from GET /v1/products (the prices rows whose mode is this mode, or video_edit with reference videos) instead of hard-coding them; billing.frozen on the retrieved task is what counts. All Kling 4.0 prices: Kling 4.0 → Pricing.

Errors

Create request errors: the request fails with an HTTP error and no task is created. Specific to this mode:

Other errors of the create endpoint (401, 402, 429 and so on): Create a task → Errors.

Task failures: the create request succeeded (200) and the task later turned failed. There's no HTTP status: the reason is the task's error.type, from retrieving the task or the webhook.

Failure reasons shared by every mode (such as a balance too low for the task): Kling 4.0 → Errors; all of them: Failure reasons.