KLING AI: Launches Kling Image 4.0, Opens API to Developers

On 17 November 2026, Kling AI launched its Image 4.0 model with native 8K and 2,000 characters per image, and opened API access to enterprises and developers.

NewTimeSpace News: On 17 November 2026, Kling AI launched its new image creation model Kling Image 4.0 and opened API access to enterprise customers and developers. The model uses a unified understanding-and-generation architecture that brings visual understanding and image generation into one system, linking intent comprehension to visual output. The company says realism, text rendering, instruction following and consistency have all improved.

The release groups the upgrades into three areas. For realism, physical rules and visual perception are incorporated into generation, with work on skin texture, material detail and light transitions, plus first-time support for native 8K output that leaves room for display and cropping. For text, the model can place 2,000 characters in a single image, rendering small type more clearly and handling multi-block layouts; it ships with more than 70 preset commercial fonts and allows uploaded font references. For series, a single prompt can produce up to 10 coherent images sharing one creative brief, and the model supports long-format e-commerce detail pages and web design layouts.

According to the official release note, the text input ceiling is 15,000 tokens, up to 10 reference images can be supplied, output supports native 2K, 4K and 8K, and languages covered include Simplified Chinese, Traditional Chinese, English, Japanese, Korean, Spanish and Portuguese. More than 110 preset grading styles are built in, and a native agentic capability helps plan the content and emphasis of each image in a series.

The API update notice lists /image-generation/kling-4.0 for image generation and adds /long-image-generation/kling-4.0 for long images, where e-commerce detail pages require a product image and offer three lengths. Series size can be set through settings.series_amount at 2 to 10 images or left to auto. Billing has changed from output-only charging to charging per image for both input reference images and outputs, with output unit prices varying by resolution.

×
Share to WeChat

Open WeChat, use the "Scan", and share to my Moments.