On this page
On November 17, 2026 (the official changelog lists the calendar date only without a specific timestamp or timezone), Kling AI announced the availability of the Kling Image 4.0 API, introducing two dedicated endpoints for standard image generation and long-form vertical visual layouts, headlined by native 8K output, 2-to-10 image series generation, and specialized e-commerce and webpage design skills.
According to the official llms.txt documentation index, Kling Image 4.0 is maintained as a separate product line from the Kling 4.0 video model released on October 15, 2026 (covered in our Kling 4.0 video keyframes and reference assets guide). Similarly, the November 14, 2026 changelog entry adding 4:3 and 3:4 aspect ratios applies to the video line rather than this image API release.
Dual API Endpoints and Standard Generation Specs
Kling Image 4.0 splits task submission across two endpoints while sharing GET /tasks for asynchronous status polling:
Standard Image Generation (POST /image-generation/kling-4.0): **Supports both Text-to-Image and Image-to-Image workflows. The **settings.result_type parameter defaults to single (controlled by settings.n, defaulting to 1 within a range of 1 to 9), or can be set to series (where n is ignored in favor of series_amount, defaulting to 4 and accepting integers from 2 to 10 or the string auto; auto dynamically selects between 2 and 10 images based on inputs and consumes concurrency according to the actual output count).
Resolution and Aspect Ratio Options: **The **settings.resolution parameter defaults to 2k and supports 2k, 4k, and native 8k across both single and series modes (the changelog also notes standard or adaptive options). settings.aspect_ratio defaults to auto and supports 16:9, 9:16, 1:1, 4:3, 3:4, 3:2, 2:3, 21:9, and auto. Optional watermarking via options.watermark_info.enabled defaults to false.
Reference Inputs and Prompt Limits: **Every request requires a non-empty prompt inside **contents. Combined reference images (refer_image) and reference elements (element) are capped at 10 items total (referenced via @id in prompts, with a 30 MB per-image size limit; the prompt itself does not count toward the 10-item cap). Prompts may contain up to 50,000 English characters or 21,000 non-English characters, with mixed-language inputs governed by English_chars/50000 + non_English_chars/21000 <= 1.
Long-Image API (long-image-generation) Constraints
The dedicated POST /long-image-generation/kling-4.0 endpoint differs significantly from standard image generation in output limits and input schemas:
One Output Per Request and 1K/2K Cap: **Each request produces exactly **one full vertical image and does not accept an output count parameter. settings.resolution defaults to 1k and supports only 1k and 2k (native 8K is not supported). In addition, type=element library references are not supported on the long-image endpoint.
E-Commerce Detail Pages vs. Webpage Visual Design: **Requests must specify **settings.skill as either ecommerce or webpage. The ecommerce skill is Image-to-Image only, requires 1 to 3 refer_image inputs, disallows style_refer_image and subject_refer_image, and uses settings.length (defaulting to medium, with short, medium, and long options). Conversely, webpage supports both Text-to-Image and Image-to-Image, ignores settings.length, and uses style_refer_image and subject_refer_image rather than refer_image.
Pricing Structure: Both Reference Inputs and Outputs Are Billed
Per the official image pricing schedule (pricing/base/image.md), unlike earlier Kling image models that billed solely for generated outputs, Kling Image 4.0 charges for both input reference images/elements and output images under the formula: Total Cost = (Reference images + Reference elements) x Input unit price + Output images x Output unit price. While the API changelog summary mentions only reference images, the pricing table explicitly counts reference elements at a 1:1 ratio with reference images; note also that although the long-image formula row in the pricing table still displays the word "elements," the long-image API itself does not accept type=element inputs.
Input Rate (Shared Across All Modes): ****0.8 Units ($0.003) per reference image or reference element (1 reference element = 1 reference image).
Standard Single & Series Output Rates: **2K costs **16 Units ($0.063) per image; 4K costs 24 Units ($0.094) per image; native 8K costs 40 Units ($0.157) per image.
Long Image · Webpage Visual Design Output Rates: **1K costs **140 Units ($0.549) per image; 2K costs 152 Units ($0.596) per image.
Long Image · E-Commerce Detail Page Output Rates: **Short 1K costs **60 Units ($0.235) per image and Short 2K costs 72 Units ($0.282) per image; Medium 1K costs 120 Units ($0.470) per image and Medium 2K costs 140 Units ($0.549) per image; Long 1K costs 200 Units ($0.784) per image and Long 2K costs 220 Units ($0.862) per image.
Official Worked Example: **A Kling Image 4.0 task using 2 reference images and 1 reference element to generate a 4-image 2K series incurs $0.003 x 3 ($0.009) in input fees plus $0.063 x 4 ($0.252) in output fees, totaling **$0.261. Regional availability restrictions, SLA terms, concurrency quotas, and any open-weight or Hugging Face releases have not been announced in the official documentation.
