Film · Model Watch
Kling 4.0 is live. Now comes the 30-second test.
Longer AI-generated shots are exciting. Keeping the characters, action and camera under control from beginning to end is what filmmakers really need.
Reporting updated Sep 29, 2026
THE TAKEAWAY
Kling 4.0 has officially launched, giving filmmakers the ability to generate up to 30 seconds of video in a single take, with more control over characters, environments and the moments that shape a scene. The exciting part isn't simply making a longer video. It's having more ways to direct what happens inside it.
WHAT HAPPENED
On October 15, Kuaishou officially launched Kling 4.0, moving the model beyond its limited preview period.
The release brings together longer generations, expanded visual references, multiple keyframes and more sophisticated camera direction.
One early demonstration shows why that matters.
Johnson Sheng, a senior creative director at Ogilvy and Kling creative partner, used two character references and a single environment image to create a 30-second fight sequence. Characters smash bottles, leap over furniture and interact with objects throughout the room.
According to Sheng, the entire sequence was generated as one continuous, unedited take.
He even acknowledged one imperfect moment:
“The only minor hiccup is the eye-poking move at the end.”
It's a useful glimpse of what filmmakers can attempt when they have enough duration to stage a scene rather than assemble a collection of disconnected shots. Sheng's demonstration was produced during Kling's early-access program, not as an independent benchmark.
WHY IT MATTERS
Imagine directing an action sequence. Your character enters a room, crosses behind furniture, picks up an object and turns toward the camera.
A beautiful opening frame isn't enough. The character needs to remain recognizable. The object must stay where it belongs. The camera has to follow the action without losing its place in the scene.
That's where longer generation and reference control become interesting.
Kling is giving creators more ways to communicate those decisions before the shot is generated. And with the full model released, filmmakers can begin testing those capabilities in their own projects.
Some announced features are still arriving. Kling's current documentation identifies 4K/10-bit HDR output and two-minute video extensions as forthcoming, while enterprise API access is also planned.
FOUR TOOLS FOR DIRECTING A SHOT
30-second generations
Give action, dialogue and camera movement room to develop without an immediate cut.
15 reference inputs
Combine images, video, subjects and voice references to define characters and environments.
10 keyframes
Specify important poses, transitions and story moments throughout a sequence.
Agent-driven workflows
Kling 4.0 now integrates with Kling MCP and CLI, allowing supported AI-agent workflows to access generation controls.
VEYR ANGLE
When VEYR first covered Kling 4.0 in September, the full model was still waiting for release.
Now the conversation changes.
The measure of progress isn't how spectacular one generated clip looks. It's whether creators can make a sequence they imagined, revise it deliberately and carry those decisions into the next shot.
Thirty seconds is a bigger canvas. The interesting question is how much creative control filmmakers can exercise over every second.

