Kling AI’s 4.0 video model supports clips up to 30 seconds, with reference inputs and keyframes intended to give creators more control over characters, action and camera movement. The shift is from making a convincing shot to helping shots hold together as part of a story, an important test for AI video beyond the demo reel.
Watch Desk analysis
What happened
TechNode reports that Kling 4.0 can generate a video up to 30 seconds in one go, accept up to 15 multimodal reference assets, and use up to 10 keyframes. References can include images, video and audio. The two controls do different jobs: references guide how a video should look and feel, while keyframes let creators specify character states, action changes and scene transitions.
The model can also extend existing videos. Kling’s spokesperson told TechNode that longer-form generation still faces challenges with stability, controllability and narrative coherence. A longer clip is not, by itself, a coherent story. A small distinction, perhaps, except for anyone hoping the model will direct the film as well as render it.
Why it matters
For creators, references and keyframes offer more specific ways to guide a generation than text prompts alone. If they work reliably, they could mean fewer rounds of generating clips and sorting through near-misses. The bigger prize is continuity: characters, camera direction, action and style that remain consistent across a sequence.
That matters as AI video tools move from isolated clips towards adverts, short films and other edited work. The evidence here is a product description and the company’s account of its aims, not proof that the model can deliver dependable long-form filmmaking.
Our read
The useful change is not simply a longer maximum clip. It is the attempt to give creators more control over what happens within that clip and how a story develops. That is a more practical ambition than chasing another impressive-looking few seconds. The real test is whether creators can repeat a desired result, keep continuity across shots and spend less time generating footage they will never use.
What to watch
- How well references and keyframes preserve character and scene continuity in complex generations.
- Whether video extension maintains action, camera direction and audio-visual synchronisation.
- Whether creators find the controls reduce revision work, rather than just adding more settings to manage.
Discussion spark: For AI video to become a useful production tool, which matters more: precise control over individual shots or reliable continuity across an entire story?
Sources and evidence
- Source update (16 October 2026, 08:34 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.