Kling 3 API Pricing and Integration Guide in 2026: Budgeting, Async Jobs, and Production Patterns

• Kling 3.0 text-to-video and Kling 3.0 image-to-video routes are present
• the catalog lists per-second billing
• the catalog lists a 3 to 15 second billable range
• the public API references document async task creation through POST /v1/videos/generations
• task lookup is documented through GET /v1/tasks/{task_id}
• generated output links are documented as time-limited, so assets should be saved promptly

TL;DR

• The current repo catalog lists Kling 3.0 at $0.075 per second on EvoLink.
• The same catalog lists a minimum billable duration of 3 seconds and maximum billable duration of 15 seconds.
• The public API reference supports both text-to-video and image-to-video task submission.
• The safer production pattern is submit job -> store task ID -> poll task -> persist output immediately.

What is verified right now

The same catalog snapshot lists:

• minimum billable seconds: 3
• maximum billable seconds: 15

Quick cost table

That estimate is helpful for planning, but it is not the whole production cost. Real workloads should also budget for:

• failed jobs
• retries
• test generations
• storage or post-processing outside the generation request itself

When Kling 3 is the right choice

Based on the current repo descriptions, Kling 3 is a strong fit when you need:

• a 3 to 15 second clip range
• both text-to-video and image-to-video
• support for single-shot and multi-shot patterns
• a programmable route that fits task-based backend orchestration

The image-to-video route is especially useful if your workflow starts with a locked first frame, an end frame, or a scene-control asset rather than a pure prompt.

How the integration flow works

The current public API references in this repo document a standard task-based workflow:

1. POST /v1/videos/generations to create the generation task
2. store the returned task_id
3. GET /v1/tasks/{task_id} to query progress and result
4. save the final asset before the returned link expires

Example request

curl https://api.evolink.ai/v1/videos/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kling-v3-text-to-video",
    "prompt": "A cat running on a grass field under bright sunshine",
    "duration": 5,
    "aspect_ratio": "16:9",
    "quality": "720p"
  }'

Production advice that matters more than the code sample

1. Treat the task ID as a first-class object

Do not treat the create request as disposable. Persist:

• task_id
• normalized request payload
• user and job metadata
• timestamps
• final output location

If a webhook or polling worker fails later, that task ID is what lets you recover state cleanly.

2. Save completed assets immediately

The current public API references for these video routes document output links as time-limited. That means your pipeline should mirror or persist the final asset quickly instead of assuming the returned URL is durable forever.

3. Start with the shortest acceptable clip

Per-second billing makes short test clips the simplest way to control waste. For early QA:

• start with 3s or 5s
• use 720p unless the product requirement truly needs more
• only move to longer durations after prompt quality is stable

4. Separate prompt testing from production generation

Video prompts are expensive to iterate compared with text prompts. A useful pattern is:

• small internal test set for prompt QA
• approval or guardrail step
• production submission only after prompt structure is stable

When to compare Kling 3 with another route

Do not force Kling 3 to be the answer to every video job.

Use Kling 3 when its documented strengths match the job:

• prompt-first short clips
• image-guided motion from a controlled first frame
• multi-shot structures where you want task-based orchestration

If your workflow depends on another provider's native feature set, fixed-duration presets, or a different billing shape, compare routes based on the real job rather than brand preference.

Pre-launch checklist

• Verify the current live price on the route page before publishing a quote.
• Confirm whether your workflow needs text-to-video or image-to-video.
• Persist task_id for every submitted job.
• Save outputs before temporary links expire.
• Add retry and failure handling around task polling.
• Test billing with your real clip-duration distribution.

FAQ

What is the current Kling 3 API price on EvoLink?

What duration range is documented for Kling 3?

Does Kling 3 support both text-to-video and image-to-video?

Is Kling 3 synchronous or asynchronous?

What endpoint is used to create a Kling 3 video task?

What is the safest production pattern for Kling 3 integration?