Kling 4.0 Flash Reference to Video API

Send up to 10 reference images in image_urls and up to 5 reference videos in video_urls to POST /api/v1/videos/generations with model: kling-4-0-flash-refs. Images set the style, scene or subject; videos show motion and camera work to imitate. You get a 720p clip of 3 to 20 seconds with native audio; track it with the returned img_uuid.

Get API Key

Storyboard table with photos and a tablet playing a dance clip

Show the Model What You Mean

Name assets in the prompt

Write <<<image_1>>> or <<<video_1>>>, numbered from 1 in array order. A number with no matching asset is rejected.

Read the Kling 4.0 Flash Reference to Video API docs

Numbered photo prints laid out in a row on a desk

Reference videos for motion

Up to 5 videos, each at least 3 seconds and together at most 20 seconds, guide motion and camera work. They are not edited.

Read the Kling 4.0 Flash Reference to Video API docs

Dancer silhouette with motion trails in a studio

Images-only requests cost the same as Flash

Without videos you pay 5.5715 tokens ($0.11143) per generated second. Reference images are never charged.

Read the Kling 4.0 Flash Reference to Video API docs

Stack of photo prints next to a small coin jar

At least one image or video is required.

Kling 4.0 Flash Reference to Video Request Fields and Limits

model

Always kling-4-0-flash-refs.

prompt (required)

Up to 2500 characters. Refer to assets as <<<image_N>>> / <<<video_N>>>.

image_urls

Up to 10 reference images. Up to 10MB each, width and height at least 300px, aspect ratio between 2:5 and 5:2.

video_urls

Up to 5 reference videos, each at least 3 seconds and at most 20 seconds in total. MP4 or MOV, up to 200MB, height 700-4553px.

duration, resolution and ratio

duration: integer 3-20, default 5. resolution: 720p only. ratio: 16:9 (default), 9:16, 1:1 or 21:9.

Rejected fields

image_url / image_end_url / keyframes, reference audio, sound / generate_audio, negative_prompt, seed, model_params.multi_shot and 1080p / 4K are rejected before any tokens are charged.

custom_callback_url

Optional public HTTPS URL. GoEnhance POSTs the same body as GET /api/v1/jobs/detail when the task status changes, with up to 3 retries.

Two rates, depending on whether you send reference videos.

Kling 4.0 Flash Reference to Video API Pricing

Images only: 5.5715 tokens ($0.11143) per output second

Example: 5 seconds costs 27.86 tokens ($0.5572).

With reference videos: 3.7143 tokens ($0.074286) per billed second

Billed seconds = output seconds + max(total reference video seconds, output seconds). Video length is measured before charging and rounded up.

5 seconds with an 8-second reference video

13 billed seconds: 48.29 tokens ($0.9658).

10 seconds with a 5-second reference video

20 billed seconds: 74.29 tokens ($1.4858).

Integrate the Kling 4.0 Flash Reference to Video API

1. Create an API key

Generate a key at /keys and send it as Authorization: Bearer <key>.

2. Prepare the request

Upload your references to public URLs and list them in image_urls / video_urls.

3. Store img_uuid

POST model, prompt and the references to /api/v1/videos/generations, then poll GET /api/v1/jobs/detail with the img_uuid.

4. Save the video

When the status is success, download the video and store it on your side.

Create an API Key

Consistent subjects and borrowed motion.

What Developers Build with Kling 4.0 Flash Reference to Video

Character-consistent clips

Keep the same person, pet or mascot across scenes with reference images.

Dance and motion templates

Pair a user's photo with a reference dance video to copy the movement.

Product placement

Put a product from <<<image_1>>> into a scene from <<<image_2>>>.

Camera move presets

Reuse a reference video's camera path on new subjects.

Common questions about requests and billing.

Kling 4.0 Flash Reference to Video API FAQ

Do I need both images and videos?

No. At least one image or one video is required.

Are reference videos edited?

No. They guide motion and camera work; to change an existing clip use kling-4-0-flash-edit.

How are reference videos billed?

At 3.7143 tokens ($0.074286) per second of output + max(reference video seconds, output seconds).

Can I add a first frame?

No. image_url is rejected here; use kling-4-0-flash for a first frame.

Is audio included?

Yes. Every clip has native audio and the audio switches are rejected.

Start Building with the Kling 4.0 Flash Reference to Video API

Create an API Key