Kuaishou's Kling AI team officially launched the Kling 4.0 API on October 22, 2026, gradually opening it to enterprise customers and developers worldwide. The launch is the latest step in the AI video race, where longer, more controllable generations are the new battleground — and third-party platforms are already racing to integrate it.

The headline capabilities are aimed squarely at creators who want director-level control. Kling 4.0 generates native single takes of up to 30 seconds, supports up to 10 keyframes per generation for shot-level control, and offers Omni Reference with up to 15 combined reference inputs across images, video, voice and subjects. Audio gets an upgrade too, with native two-channel stereo audio and tighter lip-sync.

Adoption is already underway: AI video platform SelfyzAI announced on October 23 that it is among the first platforms to integrate the Kling 4.0 API, after taking part in the model's first-batch internal beta review ahead of release. Its workspace model picker now offers Kling 4.0 alongside existing models including Seedance 2.5, so creators can match each project to the model best suited to it without changing how they work.

For video creators, the practical takeaway is control over length and consistency: 30-second native takes reduce the stitching and looping that shorter models force, while keyframes and multi-reference inputs aim at keeping characters and scenes consistent across a shot. Pricing and wider availability details are still rolling out as the API opens gradually — check Kling's official pages before building on it.

Sources