Skip to content

Models & products

Kling Image 4.0 launches with 8K output and 10-image sets

Edited and reviewed by AI Ping Editorial Team

3-line summary

  1. Kling AI launched its new image model on November 17.
  2. Long page concepts have different output and resolution limits.
  3. Provider documentation lists resolution-based rates and reference-image charges.
PracticalDesignersMarketers
Three imagined prints showing different views of the same blue teapot
AI-generated conceptual image: three product views, generated with Codex · AI Ping

Kling AI launched Kling Image 4.0 on November 17, announcing native 8K output and coordinated sets of up to 10 images from one prompt. It targets film, advertising and design work. Distribution routes are also available, with model documentation and generation listings on fal, Runware and each::labs.

The set feature targets connected assets such as product angles or storyboard shots. The fal API distinguishes independent images from a coherent series. A series can contain two to 10 images, or let the model choose, and returns them in order. Independent-image mode supports up to nine outputs in a separate request setting.

Kling says the model can place up to 2,000 characters in an image, with more than 70 preset fonts and font-reference images. It lists Korean, English, Japanese, Spanish, Portuguese, Simplified Chinese and Traditional Chinese. Improvements to small type and multi-section layouts are company claims.

An imagined leaf poster with bars representing text placement
AI-generated conceptual image: a poster illustrating detail and text layout, generated with Codex · AI Ping

Different conditions for images and long page layouts

The 8K label depends on aspect ratio. Runware documents square 8K output at 8,192 by 8,192 pixels and 16:9 output at 10,880 by 6,144. Regular image generation offers 2K, 4K and 8K options, with different pixel dimensions for a square product image and a widescreen composition.

Long product pages and webpage designs follow a separate workflow. On each::labs, the output is one tall image at 1K or 2K. E-commerce layouts require one to three references; webpage concepts can start from text. The documentation describes an image concept, rather than a functioning website or editable page document.

A long printed concept layout featuring an amber bottle
AI-generated conceptual image: a printed long product-page concept, generated with Codex · AI Ping

The each::labs text-to-image listing prices each output at $0.063 for 2K, $0.094 for 4K and $0.157 for 8K. Multiplying those published rates by 10 gives $0.63, $0.94 and $1.57 for 10 outputs. These are calculations using that provider’s rates; cost varies with resolution and image count even within one request.

An independent assessment measures something different from the advertised specifications. The Artificial Analysis overall text-to-image table opened on November 18 listed Kling Image 4.0 seventh, with Elo 1,180, 5,550 samples and a 95% interval of 1,167–1,193. Its pairwise human-preference votes do not measure whether every character in a 2,000-character image is correct.

Access also depends on the interface. The simplified generator on fal’s introduction page produces one 2K image and accepts one reference. Higher resolutions, sets and multiple references require its playground or API. Free runs depend on the current offer and eligibility; when that allowance is unavailable, the page directs users to sign in and use fal credits.

Long-image rates are listed separately. With one reference, each::labs prices a short e-commerce concept at $0.238 in 1K or $0.285 in 2K. Its default medium-length 1K configuration costs $0.473. Webpage concepts are listed at $0.549 for 1K and $0.596 for 2K. Each additional reference adds $0.003.

Runware likewise separates output charges from input-image costs, listing $0.003 per reference. Its page-design mode limits output to 1K or 2K. The e-commerce length setting controls the number of content sections.

Sources