Kling 4.0 does 30 second takes with 15 references

Kling AI released Kling 4.0, built around longer single takes and much heavier reference control. The headline capability is native 30-second generation in a single generation, without stitching, aimed at long takes and continuous storytelling.

The reference system is where the control lives. Omni Reference accepts up to 15 reference assets per generation: up to 10 images, 5 videos totalling 30 seconds, across 7 subjects, and a subject can carry a specified voice. Separately, up to 10 keyframe images give precise control of pose, expression, transitions and story beats. Prompt complexity goes up to 8,000 tokens with automatic expansion and refinement.

Audio is two-channel stereo with improved lip sync accuracy, and the model adds precise editing of expressions, body movement, camera position, style and background. Supported languages are Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, with dialect and accent variants including Beijing, Taiwanese, Northeastern Chinese, Sichuanese and Cantonese, plus American, British and Indian English. Multilingual text, emoji and logo rendering are supported. A 21:9 ultrawide aspect ratio is available.

Two headline capabilities are not shipping now. 10-bit HDR at 4K and 1080p and video extension up to 2 minutes are both marked coming soon, so the 4K HDR ceiling in the announcement is a roadmap item rather than an available output today.

Availability is the other constraint. Kling 4.0 is “entering early access, with a wider rollout coming in October,” and Kling 4.0 Flash is limited to Ultra Yearly subscribers from September 28. No pricing or per-generation credit cost is published.

(Original source: Kling AI)