Kling 4.0

Kling 4.0 is pitching itself at people who want more control out of AI video than the usual prompt-and-pray routine. The model is aimed at creators, marketers and studios, and the pitch is blunt: lock the story first, then let the machine fill in the rest.

The control comes from a fairly packed input stack. Users can give it up to 10 keyframes, 15 reference assets and prompts as long as 8,000 tokens, and Kling says that can produce 30-second clips in a single pass. That is a very different promise from the common reroll loop, where the output keeps drifting until someone gives up or gets lucky.

Audio is part of the package too. Kling 4.0 includes stereo sound and lip sync in more than 10 languages, which makes it more than a silent image engine with ambitions. The company is also pushing a feature called Creative Recreation, which can swap products or characters inside existing videos.

There is also a faster, cheaper route for drafts. Kling 4.0 Flash is meant for quick iteration, while MCP and CLI access let the model slot into agent workflows. That combination suggests the target audience is not just people making one-off clips, but teams trying to stitch video generation into a production pipeline.