How to Use the Kling 4.0 API

Nov 4, 2026

EmpirioLabs AI

Kling 4.0 is available on EmpirioLabs. It is Kuaishou's newest Kling video model, launched on October 15, 2026. One request makes a clip of 3 to 30 seconds at 720p or 1080p with native audio, and the inputs you attach decide what it does: a prompt alone, keyframes, reference images and videos, or a video to edit.

What the model does

Text to video. A prompt of up to 26,000 English characters, at 16:9, 9:16, 1:1 or 21:9.

Keyframes. A first frame (image), middle frames (middle_images) and a last frame (image_end). The clip passes through them in order and follows the first frame's aspect ratio.

References. Up to 10 images in total for characters, objects, scenes or style (reference_images), and reference videos for motion, camera work or style (reference_videos). Refer to them in the prompt as @image_1, @image_2 and @video_1.

Video editing. Send a clip of 3 to 30 seconds as video with an instruction in the prompt. The result keeps the clip's length and aspect ratio.

How to call Kling 4.0

Video generation is asynchronous: submit the job, then poll it. Set model to kling-4-0:

curl -X POST https://api.empiriolabs.ai/v1/videos/generations \
  -H "Authorization: Bearer $EMPIRIOLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-4-0",
    "prompt": "A hummingbird hovers at a red flower in slow motion, wings shimmering, garden ambience.",
    "resolution": "720p",
    "duration": 5
  }'
# {"job_id": "...", "status": "processing", "poll_url": "/v1/jobs/..."}

curl https://api.empiriolabs.ai/v1/jobs/JOB_ID \
  -H "Authorization: Bearer $EMPIRIOLABS_API_KEY"

The finished job returns a signed video URL, the clip's length, and the usage it was billed for. To edit a clip instead, send "video": "https://..." and describe the change, for example "prompt": "Turn the scene into a snowy winter evening".

Operational notes

The defaults are 1080p and 5 seconds. A video edit ignores duration and returns the length of the clip it edits.
Video inputs must be 700 to 4553 pixels on each side, so a 480p clip is too small to edit or to use as a reference. Images need at least 300 pixels and an aspect ratio between 2:5 and 5:2.
Any video input, whether a clip to edit or a reference video, is billed at the video input rate on at least twice the length of the output. The job result reports the billed seconds.
In the Playground, one attached image becomes the first frame, two become the first and last frames, and three or more are used as references. Over the API, the field you put each file in sets its role.

Pricing

Kling 4.0 is pay as you go, billed per second at a rate set by the resolution, with a separate rate for requests that include a video. Current rates are on the model page and the pricing page.