Kling 4.0 Brings 30-Second Storytelling and 10-Keyframe Control to AI Video Race

Kling 4.0 is Kuaishou Technology's next-generation AI video generator, scheduled for official launch in October 2026, that generates clips of 3 to 30 seconds with long takes, continuous storytelling, up to 10 keyframe images and up to 15 multimodal references.

Kuaishou Technology's Kling AI will launch Kling 4.0 in October, bringing native 30-second video generation, 10-keyframe narrative control and up to 15 multimodal references to a market that ByteDance is contesting with its own generative video models. Kling AI confirmed the release window in an official announcement and opened a limited beta of Kling 4.0 Flash to select subscribers on Sept. 28, 2026. The release positions Kuaishou's video-generation line around long-form, director-style storytelling rather than short looping clips, and it gives paying subscribers early access to a model the company says delivers stronger visual realism, creative control and narrative completeness.

The October launch is part of a period of rapid iteration in Chinese AI video generation. Kuaishou, which operates the short-video platform of the same name, has used Kling as a standalone creative tool and an extension of its content ecosystem. ByteDance, owner of Douyin and TikTok, has pushed competing video models, making output length, character consistency and directorial control the central battleground. Kling 4.0's feature list is aimed directly at those criteria, and the company's decision to preview the model through a Flash beta gives it a chance to gather production-scale feedback before the general release.

What is Kling 4.0 and when will it launch?

Kling AI described Kling 4.0 as the all-new version of its video generator, with a launch window of October 2026. The company published the timeline in release notes on its official site and repeated it on its X account. The release notes state:

The all-new Kling 4.0 will officially launch in October. As part of the Kling 4.0 lineup, Kling 4.0 Flash became available to a limited group of users for early access on September 28.

Kling AI's official X account echoed the timeline: Kling 4.0 is coming in October, and Kling 4.0 Flash is live now for Ultra Yearly subscribers. The same post added the tagline the company uses across its marketing: You call the shots!

How does the Kling 4.0 Flash beta fit into the launch?

Kling 4.0 Flash is an early-access tier within the Kling 4.0 lineup, not a separate product. Kling AI said it became available Sept. 28 to a limited group of users, including Black Gold and Ultra Yearly subscribers. The release history page on Kling AI's site logs that date as a major update under the heading Meet All-New Kling 4.0! You Call the Shots.

The beta gives Kuaishou a way to test real-world workloads before the full October launch. Early Flash access has already demonstrated strong cinematic storytelling and character consistency, according to Kling AI's description of the version, signaling that the company considers the model stable enough to expose to paying subscribers before general availability. The gating of access to top-tier subscribers also makes the beta a retention incentive for Kuaishou's highest-paying users.

What technical capabilities does Kling 4.0 introduce?

Kling 4.0 supports single-generation video clips from 3 to 30 seconds, a range that accommodates short-form social posts and longer narrative scenes. The company said the model supports long takes and continuous storytelling, implying that a single generation can maintain scene coherence rather than stitching together isolated clips. That distinction is significant in an AI video market where many models produce brief, disconnected shots that require heavy editing to assemble into a narrative.

For narrative control, Kling 4.0 accepts up to 10 keyframe images as input. Creators can use those frames to fix shot composition, camera movement and story beats before generation. In the same task, the model accepts up to 15 multimodal references: up to 10 images, five video clips and seven subjects or elements. That combination is designed to preserve identity and style across a scene or a sequence, addressing one of the most persistent complaints about AI video generation: characters and objects drifting between shots.

Kling 4.0 also expands output formats. It supports 720p, 1080p and 4K resolution, with 1080p and 4K listed as coming soon, and adds 10-bit HDR for the higher resolutions. A 21:9 ultrawide aspect ratio gives creators a cinematic framing option beyond standard vertical and horizontal formats. The combination of higher resolution, HDR color and ultrawide geometry points to professional production workflows rather than mobile-only social clips.

Audio is part of the update. Kling 4.0 supports two-channel stereo output and improved lip-sync accuracy, an area where many video-generation models remain weak. The combination of synchronized dialogue and ultrawide visuals suggests Kling 4.0 is aimed at narrative filmmaking, advertising and other production use cases, not just social clips. Stereo audio also matters for interviews, dialogue-driven scenes and any content where voice clarity is central.

Text input has also grown. Kling 4.0 increases prompt length to up to 8,000 tokens and supports multiple languages, accents and dialects. Longer prompts let creators specify camera moves, lighting, subject behavior and dialogue in a single instruction set, while multilingual support lowers the barrier for non-English-speaking users. The model's ability to handle accents and dialects is a practical advantage for global creators who want regionally specific speech.

Kling 4.0 announced specifications

Capability	Specification
Video length	3 to 30 seconds per single-generation clip
Keyframe input	Up to 10 images
Multimodal references	Up to 15 per task: 10 images, 5 video clips, 7 subjects or elements
Output resolution	720p now; 1080p and 4K coming soon
Color and contrast	10-bit HDR at 1080p and 4K
Aspect ratio	21:9 ultrawide
Audio	Two-channel stereo with improved lip-sync accuracy
Prompt length	Up to 8,000 tokens
Languages	Multiple languages, accents and dialects

How will Kling 4.0 change what creators can produce?

Kling AI said Kling 4.0 brings major improvements in visual realism, creative control and narrative completeness. The three-part framing matters because each maps to a known limitation of earlier video models: realism has improved across the industry, but control and narrative structure have lagged. Kling 4.0's keyframe system and multimodal references are the mechanisms meant to close that gap.

For a creator, the practical workflow starts with a prompt and ends with a 30-second, audio-synced clip. Kling AI's announced controls support a step-by-step production process:

1. Compose a prompt of up to 8,000 tokens, using natural language in supported languages, accents or dialects.
2. Upload up to 10 keyframe images to define shot composition, camera movement and narrative beats.
3. Add up to 15 multimodal references, including up to 10 images, five video clips and seven subjects or elements, to maintain character and style consistency.
4. Select output parameters, including 720p today or 1080p and 4K once available, and choose 21:9 ultrawide framing for cinematic composition.
5. Generate a clip of 3 to 30 seconds with two-channel stereo audio and monitor lip-sync accuracy in the result.

The multimodal reference count is notable because it lets a creator separate the elements that must stay consistent. Up to seven subjects or elements can be defined independently, so a character, an object and a location can each be specified without forcing one prompt to carry the entire burden. Five reference video clips can define motion, pacing or style, while 10 reference images can lock down framing and appearance. That modular approach is a direct answer to the consistency problems that have limited AI video production.

What does the launch mean for the AI video market?

Kling 4.0 enters a market defined by Chinese platform companies pushing model quality upward. ByteDance has made generative video a priority, and Kuaishou's Kling line is its direct competitive response. The October release date, coming shortly after the Flash beta, keeps Kuaishou on an aggressive release cadence and signals that the company intends to hold its position in the frontier of video generation rather than cede the high end to ByteDance or Western competitors.

For subscribers, the gating of Flash access to Black Gold and Ultra Yearly subscribers is a signal about where Kuaishou expects revenue to come from. The company is treating advanced video generation as a premium, subscription-driven product, with early access used as a retention incentive for its highest-paying tiers. The September beta, the October launch and the promised arrival of 1080p and 4K output form a staged rollout that gives subscribers a reason to stay engaged across multiple releases.

What are the early signals from Kling 4.0 Flash?

Kling 4.0 Flash is the first public proof point for the lineup. Kling AI released it Sept. 28 to a limited group, and the version has already shown strong cinematic storytelling and character consistency, according to the company's description of the Flash release. Those are the exact attributes Kling 4.0 is designed to improve, so the Flash beta functions as both a preview and a stress test for the full model.

The limited beta also gives Kuaishou time to adjust before the full launch. If Flash encounters quality or reliability issues among early users, the company can refine the model before October without exposing the entire user base to a problematic release. The choice of subscribers as testers also creates a controlled feedback loop with users who are invested in the product and likely to report issues in detail.

What comes next for Kling AI and Kuaishou?

Kling 4.0's official launch is scheduled for October 2026. The company has not detailed a specific day in its announcement, instead promising the release within the month. Between the Sept. 28 Flash beta and the full launch, Kuaishou is likely to refine the model based on beta feedback and prepare the higher-resolution output modes that are listed as coming soon.

The roadmap includes 1080p and 4K with 10-bit HDR, features that are not yet available in the current release. Those additions will matter for professional users who need broadcast or theatrical-grade output. The 21:9 ultrawide format and stereo audio reinforce that orientation toward production work, and the expansion of prompt length to 8,000 tokens gives advanced users the headroom to specify complex scenes in a single pass.

Kling AI's homepage tagline, You call the shots, summarizes the product philosophy behind the release. By giving creators keyframes, multimodal references and long-form generation, Kuaishou is positioning Kling 4.0 as a tool for directed storytelling rather than a black-box clip generator. The October launch will test whether that positioning holds against ByteDance's competing efforts and whether paying subscribers judge the new controls worth the premium access they have already been offered.