Kling AI's API Pricing (2026): Cost per Second and Code

Kling AI API pricing comes down to one number: how many credits (or dollars) you pay for each second of generated video. Everything else, including resolution, start frames and retries, changes how many of those seconds you end up buying. This post gives you the per-second rate for Kling on Unicorn API, works out real budgets, compares Kling with cheaper video models, and shows the code to run a clip and check its price first.

How Kling AI API pricing works

Kling is a family of video generation models from Kuaishou. You can use it two ways: through the Kling app and website, which sell consumer subscription plans in credits, or through an API, where your code sends a request and receives a video file. The two are priced differently, and that difference causes most of the confusion you will see in forum threads about Kling AI API pricing.

API pricing for video models is almost always tied to output length. Some providers and resellers charge a flat fee per generation (for example, “one 5-second clip”), others charge per second of video. Per-second billing is easier to reason about because cost scales in a straight line: double the duration, double the bill. Resolution is the second lever. Higher resolutions cost more per second on most video models.

On Unicorn API, the current Kling model is Kling 4.0 Flash Keyframe To Video. “Keyframe to video” means you give it a starting image (the first frame, passed as first_frame_url) plus a text prompt describing the motion, and it animates from that frame. “Flash” is Kling’s naming for the faster tier of a model generation.

Its price is simple:

• 17 credits per second at 720p.
• The total is multiplied by the duration field you send, then rounded up to whole credits.
• Credits cost $0.01 each on monthly plans, or $0.015 each as top-ups that never expire.

A credit is just Unicorn API’s billing unit: one balance covers every image, video, audio and chat model, so you do not need a separate account and invoice for Kling, OpenAI, Google and the rest.

What a Kling video costs in credits and dollars

Because Kling AI API pricing on Unicorn API is linear, you can work out any clip by multiplying seconds by 17. Here is the math for common lengths at 720p, at both credit prices. Check the model page for the durations the model currently accepts.

Clip length (720p)	Credits	Cost on a monthly plan ($0.01/credit)	Cost with top-up credits ($0.015/credit)
1 second	17	$0.17	$0.255
5 seconds	85	$0.85	$1.275
10 seconds	170	$1.70	$2.55
15 seconds	255	$2.55	$3.825

Two practical notes. First, the per-second rate is the same whether your clip is short or long, so there is no discount for longer clips and no penalty for short ones. Second, the rounding happens once per output, on the total. A 17-credit-per-second model always lands on a whole number, so rounding does not change Kling costs; it matters more on models with fractional per-second rates, covered below.

You never have to do this math by hand in production. Unicorn API exposes POST /v1/models/price, which returns the exact cost of a request before you run it. The request format is in the API documentation. Wiring that call into your app means you can show users a price, enforce a budget cap, or refuse jobs that would overspend.

Why the Kling API can look more expensive than the subscription

A common complaint is that the Kling API seems to cost about double what the same clips would cost inside a Kling monthly subscription. That observation is often right, and it is not a trick. It comes from how the two products are built.

Subscriptions are priced on average use

Consumer plans are sold on the expectation that many subscribers will not spend their full allowance. The heavy users are effectively subsidised by the light ones. The per-clip price you calculate from a subscription assumes you use every credit, which most people do not.

API calls are priced per clip, every time

An API bills what you actually generate. There is no unused allowance propping up the rate, but there is also no ceiling on how much you can generate, no browser in the loop, and no daily queue juggling. You can call it from a backend, a batch script, an agent or a product your customers use.

So which should you use?

• Making a handful of clips by hand each month? The app subscription may be cheaper, and it comes with a visual editor.
• Building a product, automating content, or generating many variations? You need an API. The question becomes which API gives you the best price and the least integration work.
• Using several models (Kling for hero shots, something cheaper for drafts, an image model for start frames, a chat model for prompts)? One API key and one balance across all of them saves you a lot of account management.

When you compare Kling AI API pricing between providers, compare like with like: the same model version, resolution and duration. A “per generation” price for a 5-second clip and a “per second” price are only comparable after you multiply out.

Kling compared with other video models on Unicorn API

Kling is not the cheapest video model, and it does not need to be. The useful question is how much more it costs than the alternatives, so you can decide where it earns its price. Here is how Kling 4.0 Flash sits next to the two Grok Imagine video models at 720p.

Model	Input	720p rate	5 s at 720p	Other options
Kling 4.0 Flash Keyframe To Video	First frame + prompt	17 credits/s	85 credits ($0.85)	720p
Grok Imagine Video 1.5	Image-to-video only	4.5 credits/s	23 credits ($0.23), from 22.5 rounded up	480p (12 credits for 5 s), 1080p (40 credits for 5 s), 1 to 15 s
Grok Imagine Video 1.0	Text-to-video or image-to-video	4.5 credits/s	Not available (minimum 6 s); 6 s is 27 credits	480p, 720p, 1080p, 6 to 30 s

At the same resolution and length, Kling 4.0 Flash costs about 3.7x as much as Grok Imagine Video 1.5. Grok Imagine Video 1.5 is described as having strong motion quality at a very low cost per second, but it only animates a start frame and can be busy at peak times. Grok Imagine Video 1.0 adds text-to-video and clips up to 30 seconds at the same per-second rate, which makes it the obvious pick for long, cheap clips.

Notice the rounding effect mentioned earlier: Grok Imagine Video 1.5 at 720p is 4.5 credits per second, so odd durations produce half credits that round up. A 5-second clip lists at 22.5 credits and bills as 23. That is a rounding of half a cent per clip, so it rarely matters, but it explains small differences between your spreadsheet and your invoice.

If you came here from a search for an AI video generator cost comparison, the takeaway is this: use per-second credits as your common unit, then decide which shots need Kling. If you want the background on Kling’s text-to-video line, see our guide to the Kling AI text to video models, and for the latest Grok video model, our Grok Imagine 2.0 overview.

AI API pricing comparison beyond video

Most people researching Kling AI API pricing are also pricing the rest of their stack: something to make the start frame, and often a chat model to write or refine prompts. Searches like “open ai api pricing”, “google ai api pricing” and “moonshot ai api pricing” usually come from the same project. Here is what those pieces cost on Unicorn API, in the same credits.

Image models for start frames

Model	Maker	Price per image	Good for
Nano Banana 2 Lite	Google	4 credits	Cheap, fast 1K images and edits, up to 10 reference images, 16:9 and 9:16 among other ratios
GPT Image 1.5	OpenAI	3 (low), 4 (medium), 22 (high) credits	Strong prompt adherence and legible text, optional transparent background
GPT Image 2.6 (Economy) Text to Image	OpenAI	5 credits	Newer OpenAI image generation at a flat price
GPT Image 2.6 (Economy) Image to Image	OpenAI	5 credits	Editing an existing image into a start frame

For video work, the start frame is cheap relative to the clip. Even the high-quality GPT Image 1.5 setting at 22 credits is about a quarter of a 5-second Kling clip, and Nano Banana 2 Lite at 4 credits is close to a rounding error. Since the start frame decides composition, lighting and subject, spending a few extra credits to get it right is usually the best money in the pipeline.

Chat models for prompts and planning

Chat models are priced per million tokens (a token is roughly a word fragment). Kimi K3.1 Pro, from Moonshot, costs 160 credits per million input tokens, 800 per million output tokens, and 16 per million cached input tokens. Writing a detailed motion prompt costs a tiny fraction of a credit, so using a chat model to expand short ideas into precise Kling prompts adds almost nothing to the bill. It is called through the OpenAI-compatible POST /v1/chat/completions endpoint.

For other providers’ chat and image models, including any you are pricing from Anthropic or Z.ai, check the full model catalog, which lists each model’s current credit price.

Worked example: budgeting a month of Kling clips

Say you run a small product studio that publishes short 5-second product animations at 720p. Each finished clip needs:

1. A start frame from Nano Banana 2 Lite: 4 credits.
2. A Kling 4.0 Flash clip, 5 seconds at 720p: 85 credits.

That is 89 credits per finished clip if everything works first time. It rarely does. Assume you generate two Kling takes for every clip you keep, plus two start frames. Your real cost per kept clip is 2 × 85 + 2 × 4 = 178 credits.

Now map that onto the plans on the pricing page:

Plan	Monthly price	Credits	Kept clips at 178 credits each
Starter	$25	2,500	14
Creator	$50	5,000	28
Studio	$100	10,000	56
Pro	$200	20,000	112

Change one thing and the numbers move a lot. If you draft both takes on Grok Imagine Video 1.5 (23 credits each at 720p for 5 seconds) and then run only the chosen idea on Kling, a kept clip costs 2 × 4 + 2 × 23 + 85 = 139 credits. On the Studio plan that is 71 clips instead of 56. The draft clips are not identical to what Kling will produce, but they tell you quickly whether a start frame and prompt are worth pursuing.

If your volume is uneven, top-ups suit you better. At $0.015 per credit they cost more per credit than a plan, but they never expire, so a quiet month does not waste anything.

How to run Kling 4.0 Flash and check its price in code

Here is a complete Python script that generates a start frame with Nano Banana 2 Lite, then animates it with Kling 4.0 Flash. Create a key on the API keys page and export it as UNICORN_API_KEY.

Media models on Unicorn API run as jobs: you POST /v1/jobs, receive a job with an id, then poll GET /v1/jobs/{id} until its status is "succeeded", and read the file URL from outputs[0].url.

import os
import time
import requests

API = "https://api.unicornapi.net"
HEADERS = {
    "Authorization": f"Bearer {os.environ['UNICORN_API_KEY']}",
    "Content-Type": "application/json",
}


def run_job(model, job_input, poll_seconds=5, timeout_seconds=900):
    """Submit a media job and wait for its first output URL."""
    resp = requests.post(
        f"{API}/v1/jobs",
        headers=HEADERS,
        json={"model": model, "input": job_input},
    )
    resp.raise_for_status()
    job = resp.json()

    deadline = time.time() + timeout_seconds
    while job["status"] != "succeeded":
        if job["status"] in ("failed", "canceled"):
            raise RuntimeError(f"Job {job['id']} ended with status {job['status']}: {job}")
        if time.time() > deadline:
            raise TimeoutError(f"Job {job['id']} did not finish in time")
        time.sleep(poll_seconds)
        poll = requests.get(f"{API}/v1/jobs/{job['id']}", headers=HEADERS)
        poll.raise_for_status()
        job = poll.json()

    return job["outputs"][0]["url"]


# 1. Start frame: 4 credits with Nano Banana 2 Lite
frame_url = run_job(
    "nano-banana-2-lite",
    {
        "prompt": "A ceramic coffee mug on a wooden table, morning window light, "
                  "shallow depth of field, product photography",
        "aspect_ratio": "16:9",
    },
)
print("Start frame:", frame_url)

# 2. Animate it: 17 credits per second at 720p, so 5 seconds = 85 credits
video_url = run_job(
    "kling-4.0-flash-keyframe-to-video",
    {
        "prompt": "Steam slowly rises from the mug as the camera pushes in gently",
        "first_frame_url": frame_url,
        "duration": 5,
        "resolution": "720p",
        "aspect_ratio": "16:9",
    },
)
print("Video:", video_url)

Before running jobs at volume, call POST /v1/models/price with the same model and input to get the exact credit cost. A simple pattern is to compute the price, compare it against a per-user or per-day budget, and only then submit the job. The documentation covers that endpoint’s request and response.

To expand rough ideas into detailed motion prompts, call Kimi K3.1 Pro through the chat endpoint:

curl https://api.unicornapi.net/v1/chat/completions \
  -H "Authorization: Bearer $UNICORN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k3-1-pro",
    "messages": [
      {"role": "user", "content": "Write a one-sentence camera and motion prompt for a 5-second video of steam rising from a coffee mug."}
    ],
    "max_tokens": 200
  }'

If you prefer to try prompts by hand first, the browser playground runs the same models without code, and the agents page shows how to call them from Cursor or Claude Code.

Mistakes that inflate your Kling bill

Most overspending on Kling video has little to do with the headline rate. It comes from workflow habits.

• Iterating on the expensive model. Every Kling retry costs the full per-second rate. Fix composition in the start frame (4 or 5 credits) and test motion on a cheaper video model before committing.
• Defaulting to long clips. A 10-second clip costs twice a 5-second one. If you will trim to 4 seconds in the edit, generate shorter.
• Weak start frames. Keyframe-to-video follows the first frame closely. A muddled start frame produces a muddled clip, and you pay to find that out.
• Comparing prices that are not comparable. When you read a Kling AI API pricing table on another site, check the model version, resolution, duration and whether the price is per clip or per second. Different Kling versions and tiers are priced differently.
• Not estimating before running. Batch scripts with a wrong duration or loop count can burn a month’s credits in minutes. Check the price endpoint and set a budget guard in code.
• Buying the wrong kind of credits. If you generate steadily every month, a plan at $0.01 per credit is cheaper. If your work comes in bursts, top-ups at $0.015 that never expire waste less.

Get those right and Kling AI API pricing becomes predictable: 17 credits for every second of 720p video you decide to keep, plus a few credits for the frame it starts from.