Kling 4.0 Launches: 30-Second Videos, 15 References, and 10 Keyframes
Kuaishou(快手) just says Kling 4.0, the latest video generation model from Kling AI, launched on October 15, 2026. The update is about more than sharper images: creators can plan longer sequences, combine more visual references, and exert finer control over key moments and camera movement.
The announced features point toward a more directed workflow. Instead of repeatedly rewriting a prompt for a short clip, creators can set references, plan keyframes and camera moves, then build a more continuous story in a single generation.
What is new in Kling 4.0?
Up to 30 seconds in one generation
Kling 4.0 can generate a video up to 30 seconds long in a single pass. Kuaishou says it has improved continuity in fast action, large-scale motion, tracking shots, pans and orbiting camera moves. Longer clips may help creators avoid some of the breaks and jumps that can appear when several short generations are stitched together, especially in product showcases or scenes with a continuous emotional arc.
Up to 15 assets with Omni Reference
The upgraded Omni Reference accepts up to 15 reference assets in one task, including images, videos, subjects and voice references. The published limits are up to 10 images, five videos totaling no more than 30 seconds, and seven subjects. These inputs can guide a character’s appearance, product details, performance, camera language or voice.
Reference controls are intended to reduce visual drift across shots. They do not guarantee an exact match: results still depend on the source assets, prompt and complexity of the scene.
Up to 10 keyframes
Creators can provide up to 10 keyframe images to define character states, scene changes and other visual beats. Keyframes add control over when important images appear, which can be useful for storyboards, short narratives and action continuity.
Visual, audio and language updates
Kuaishou’s launch information lists up to 10-bit HDR output, 4K and 1080p resolutions, and a 21:9 ultrawide format. For audio, the company highlights stereo output and more accurate lip sync. Kling 4.0 also expands language, dialect and accent support, and aims to keep multilingual text, logos and emojis more stable during movement.
Feature access may vary by account, plan and rollout status. Kling’s earlier introduction described some HDR output as coming soon, so creators should check the options available in their own account.
A redesigned creation page
The updated interface brings image, video and audio references into one input area, adds grid and list views, and lets creators preview, edit and add assets on a shared timeline. A new Canvas and Agent workspace is designed to organize assets and generation steps.
Two creator posts offer examples, not official benchmarks
The two X posts illustrate possible workflows, but should be labeled as creator tests, not official demos or standardized evaluations:
- FoldingEnd A.I.: four more Kling 4.0 tests shows a pageant, a spaceship over a desert market, a ghost at a harbor and a snake trying to drink coffee. The post describes reference-image, first-frame and text-to-video examples and includes some prompts.
- Zeto: Kling 4.0 Flash vs. Seedance 2.5 describes a 720p, 20-second comparison using the same prompt and reference image, alongside estimated costs in yen. The prices are the poster’s estimates, and the test conditions and model versions are set by the poster; they should not be treated as official pricing or a general performance verdict.
Using the videos: This article links to the original posts so readers can watch them on X. The posts play in the browser, but no clear original-file download option is exposed. Before re-uploading a downloaded copy or editing it into a hosted article video, obtain the creator’s permission and follow their attribution requirements. A single demo or engagement count is not evidence that a model will lead on every task.
What does this mean for creators?
Kling 4.0 brings several controls into one workflow: longer generation supports more narrative, multimodal references help maintain character and visual consistency, keyframes define story beats, and targeted editing can reduce the need to regenerate an entire clip.
This could help with short films, ads and social content, where teams need both consistency and quick revisions. Actual time savings will depend on generation reliability, wait times, credit costs and whether the edit is good enough to deliver. A specification sheet alone cannot settle that question.
Takeaway
Kling 4.0 launched on October 15, 2026, with single-pass video up to 30 seconds, up to 15 reference assets and up to 10 keyframes, along with updates to motion, audio and the creation interface. Its central change is the amount of control creators can apply to references and story timing, beyond prompt iteration alone.
These capabilities are worth exploring for continuous shots, character-led shorts and multi-scene ads. Creators should still confirm HDR availability, account limits and actual costs in the product. Community posts can serve as examples; reusing their video files requires permission.