How to Use Kling AI: Tutorial and API Documentation Guide (2026)

TL;DR

• If your goal is how to use Kling AI for API work, start with Kling 3.0 for standard text-to-video and image-to-video.
• The current public EvoLink flow is POST /v1/videos/generations and then GET /v1/tasks/{task_id}.
• Kling video generation is asynchronous. You do not wait for the final MP4 in the initial request.
• Current public docs in this repo say generated video links are valid for 24 hours, so save results promptly.
• If you are specifically looking for Kling AI API documentation, the useful docs are the model references linked later in this article, not generic marketing pages.

What this tutorial covers

• understand how to use Kling AI through EvoLink
• pick the right Kling model family
• send a first request without guessing the endpoint shape
• find the right Kling AI API documentation for the exact route they need

• Kling 3.0 supports text-to-video and image-to-video
• Kling O3 supports text-to-video, image-to-video, reference-to-video, and video editing
• Kling O1 is the higher-control route for brand-consistent generation and editing workflows
• Kling 3.0 Motion Control is the specialized route for reference-motion transfer

Step 1: Choose the right Kling model before writing code

The easiest way to fail with Kling is to start coding before choosing the right route.

Step 2: Understand the async workflow

The public API references for Kling in this repo document an asynchronous pattern:

1. Submit a generation request to POST https://api.evolink.ai/v1/videos/generations
2. Store the returned task_id
3. Poll GET https://api.evolink.ai/v1/tasks/{task_id}
4. Save the finished asset before the result URL expires

Step 3: Send your first Kling 3.0 request

For most first-time users, text-to-video is the fastest path.

Text-to-video example

curl --request POST \
  --url https://api.evolink.ai/v1/videos/generations \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "kling-v3-text-to-video",
    "prompt": "A golden retriever running through a sunlit meadow, cinematic slow motion",
    "duration": 5,
    "aspect_ratio": "16:9",
    "quality": "720p"
  }'

What is documented in the current reference:

• model should be kling-v3-text-to-video
• the prompt limit is 2500 characters
• Kling 3.0 supports async task creation through /v1/videos/generations
• generated links are time-limited and should be saved promptly

Image-to-video example

If your workflow starts from a fixed first frame, use the image route instead.

curl --request POST \
  --url https://api.evolink.ai/v1/videos/generations \
  --header 'Authorization: Bearer YOUR_API_KEY' \
  --header 'Content-Type: application/json' \
  --data '{
    "model": "kling-v3-image-to-video",
    "prompt": "The character turns, smiles, and walks toward the camera",
    "image_start": "https://example.com/portrait.jpg",
    "duration": 5,
    "quality": "720p"
  }'

The current image-to-video reference documents these constraints:

• image_start is required
• supported formats include .jpg, .jpeg, and .png
• image size can be up to 10MB
• image dimensions must be at least 300px

Step 4: Poll the task result

Once you have submitted the job, poll the task endpoint:

curl --request GET \
  --url https://api.evolink.ai/v1/tasks/{task_id} \
  --header 'Authorization: Bearer YOUR_API_KEY'

A good production habit is to persist:

• task_id
• request payload
• user or job metadata
• final output location
• timestamps for retries and completion

Step 5: Save outputs immediately

• poll until status is complete
• copy the result to your own storage
• store the durable URL in your database

If you skip that step, a successful job can still become an operational problem later.

Where to find Kling AI API documentation

• Kling 3.0 Text-to-Video API docs
• Kling 3.0 Image-to-Video API docs
• Kling O3 Reference-to-Video API docs

Common mistakes in first-time Kling integrations

• choosing O3 or O1 before you know whether simple Kling 3.0 already fits the job
• treating video generation like a synchronous endpoint
• failing to store task_id
• forgetting that result links expire
• mixing creator-oriented pricing assumptions with API billing

Which route should you start with?

Use this short decision table:

FAQ

How do I use Kling AI for the first time?

Is this a complete Kling AI tutorial?

Where can I find Kling AI API documentation?

Which Kling model should developers start with?

Is Kling generation synchronous?

No. The current public docs in this repo document an asynchronous workflow using task creation plus task polling.

How long are Kling result links valid?

Sources

• Kling 3.0 text-to-video API reference
• Kling 3.0 image-to-video API reference
• Kling O3 text-to-video API reference
• current route pricing and model metadata as of April 8, 2026