Kling 4.0 Features: Everything New in Kuaishou's 30-Second AI Video Model

Kling 4.0 is Kuaishou's latest AI video model, launched October 15, 2026. Its headline Kling 4.0 features are native 3 to 30-second single takes, up to 10 keyframes, Omni Reference with up to 15 assets, steadier motion and camera work, two-channel stereo audio with improved lip-sync, and 21:9 framing, while 10-bit HDR 4K and 2-minute extensions remain "coming soon."

Kling 4.0 is available to use right now on Kling5 AI, the site you're reading (an independent platform, not affiliated with Kuaishou): open our Kling 4.0 AI video generator to run text-to-video, first-and-last-frame, and reference-to-video workflows right away. Below is a complete breakdown of the official specifications, release timeline, model comparisons, and developer pricing as of November 2026.

Kling 4.0 at a glance

Kling 4.0 expands maximum clip length, multimodal reference limits, and audio capabilities over previous versions, though several high-end output formats are still staged for future rollout. The table below summarizes the official specifications from Kling's documentation and distinguishes what you can use today from what is still pending.

When was Kling 4.0 released?

The official Kling 4.0 release date for all users was October 15, 2026, arriving roughly eight months after Kling 3.0 launched on February 5, 2026. Rather than dropping every tier at once, Kuaishou (HKEX: 1024) rolled out the generation suite across a four-week schedule covering early subscribers, the general web app, and API developers.

• September 28, 2026 (Flash early access): Kuaishou opened Kling 4.0 Flash—a lighter, faster generation mode—in early access for Ultra Yearly subscribers. In an official blog post dated September 30, 2026, the team confirmed that Kling 4.0 was entering early access ahead of a wider October rollout.
• October 15, 2026 (Full public launch): Kling 4.0 officially launched and rolled out to all users on the Kling AI website, both internationally at kling.ai and in China at klingai.com. In October 16, 2026 announcements via Kuaishou Investor Relations and GlobeNewswire, the company noted that Kling AI had served more than 100 million users and over 50,000 enterprise customers.
• October 22, 2026 (Partner API launch): According to reporting by PANews, the Kling 4.0 API first went live with launch partners including Higgsfield, Magnific (Freepik), OpenArt, and Meitu.
• October 28, 2026 (Full API availability): Kling's official developer changelog announced that the Kling 4.0 API was fully available to developers across three dedicated endpoints: /text-to-video/kling-4.0, /image-to-video/kling-4.0, and /omni-video/kling-4.0.

If you are already tracking how Kuaishou paces its major model upgrades, you can read our analysis of the Kling 5.0 release date for context on future release cycles.

The key Kling 4.0 features, explained

Every major upgrade in Kling 4.0 targets longer narrative continuity, tighter adherence to visual references, and integrated stereo sound. Here is what each capability does according to Kuaishou's official user guide and release documentation.

Native 30-second single-take generation

Kling 4.0 generates continuous video clips ranging from 3 to 30 seconds in a single run, doubling the 15-second ceiling of Kling 3.0. This longer window supports uninterrupted single takes and multi-beat storytelling without stitching short clips together in an external editor. Kuaishou has also announced a Video Extension feature designed to extend clips multiple times up to 2 minutes in total, though that feature is still marked as "coming soon."

Up to 10 keyframes for frame-accurate pacing

When you need strict visual continuity across a sequence, Kling 4.0 keyframes support up to 10 keyframe images in a single generation. Creators can supply a first frame, both first and last frames, or a chain of up to 10 milestone images to lock down character placement and composition with pixel-level consistency across the timeline.

Omni Reference with up to 15 assets and Creative Recreation

The Kling 4.0 omni reference system accepts up to 15 reference inputs in a single task: up to 10 images, up to 5 videos with a combined duration of 30 seconds or less, up to 7 elements (including up to 3 video elements), and voice profile references. Image inputs can include standard subject photos, three-view character sheets, grid images, storyboards, and wireframes. Video inputs can guide performance, action, camera movement, and pacing, and the model even accepts untextured 3D white-model videos and depth-map videos.

Built on top of this reference engine is Creative Recreation, a workflow aimed at product ads, social video, fashion, beauty, and consumer electronics. Creative Recreation lets you keep a reference video's shot composition, pacing, and visual style while swapping in new products, characters, or brand assets.

Precise multi-clip video editing

Instead of regenerating an entire scene from scratch when one detail looks off, Kling 4.0 includes targeted video editing. You can load up to 5 clips (including 1 primary video, with each clip lasting 3 to 30 seconds and the combined total staying at 30 seconds or less) and modify facial expressions, body movement, camera angle, visual style, or backgrounds while keeping the rest of the footage stable.

Steadier large-scale motion and camera direction

Fast action has historically caused warping in generative video, and Kuaishou specifically tuned Kling 4.0 for high-speed movement and continuous physical interactions such as martial arts fights, sports, and complex action sequences. On the cinematography side, the model better interprets camera directions including push-ins, pull-outs, pans, trucking, tracking, and orbit shots, while handling shot-size transitions and multi-shot direction more smoothly. According to Kling's documentation, the system also expands and enriches prompts "with a director's eye," and features an always-on Smart Storyboard mode with no separate toggle required.

Two-channel stereo audio, lip-sync, and voice profiles

Audio-visual synchronization is active by default in Kling 4.0 and now outputs two-channel stereo audio alongside tighter lip-sync. Speech generation covers Chinese, English, Japanese, Korean, Spanish, Portuguese, and German, plus newly added support for French and Hindi. Regional accent and dialect coverage spans Beijing, Taiwanese Mandarin, Northeastern, Sichuanese, and Cantonese, as well as American, British, and Indian English.

For custom voices, Kling 4.0 accepts voice profile references only. You cannot upload custom background music tracks or sound effect files into the generator; instead, the model synthesizes environmental sound effects and musical backing directly from your visual context and text prompt.

Ultrawide 21:9 aspect ratio, visual styles, and pending 10-bit HDR 4K

Alongside standard 16:9, 9:16, 1:1, and Auto framing, Kling 4.0 adds a native 21:9 ultrawide aspect ratio for widescreen narratives. The model also broadens its stylistic range across vlog footage, realistic documentary, clay animation, glitch art, felt textures, 3D cyber, pixel art, animation, and fantasy. For output clarity, 720p and 1080p are live today, whereas 10-bit HDR at 1080p and 4K resolution remains marked as "coming soon."

Multilingual on-screen text, logos, and 8,000-token prompts

Kling 4.0 renders multilingual on-screen text, emojis, and brand logos that stay sharp even as the virtual camera pans or tracks across a scene. To describe complex multi-shot storyboards or detailed character interactions, the web app accepts prompts up to 8,000 tokens long, while the official developer API supports prompts up to 26,000 English characters.

Redesigned creation workspace, Agent tools, and MCP/CLI support

Inside the official Kling interface, a redesigned creation page consolidates images, videos, and audio references into a single input box, paired with grid and list views, timeline editing, a node-based canvas, and an integrated Agent that can execute multi-step creation tasks. For technical teams building automated pipelines, Kling 4.0 integrates directly with Kling MCP and the CLI so external AI agents can invoke Omni Reference and keyframe-to-video tasks programmatically.

Omni Reference vs Keyframes: which entry point should you use?

Because Kling 4.0 separates its workspace into two distinct entry points—Omni Reference and Keyframes—and does not allow you to combine both in the same task, picking the right starting mode is essential before setting up your assets.

• Use Omni Reference (the default entry point) when you are combining mixed media such as character sheets, product photos, motion reference videos, 3D white-model clips, or voice profiles, or when you are running standard text-to-video prompts. It is designed for flexible multimodal blending and Creative Recreation.
• Use Keyframes when your workflow relies strictly on still images and demands pixel-level frame continuity from a starting image, between first and last frames, or across a sequence of up to 10 keyframe images. Audio and video reference clips are not used in this mode.

Regardless of which entry point you choose, Kling's official user guide highlights two practical prompting habits for Kling 4.0:

1. Write concrete visual descriptions rather than abstract buzzwords. Describe specific lighting, lens movement, textures, and physical actions instead of vague phrases like "premium look."
2. Keep negative prompts limited to subtitles and music. Because Kling 4.0 automatically enriches scenes and may add on-screen captions or background score by default, the only negative instructions you generally need in your prompt are "no subtitles" and "no background music."

How does Kling 4.0 compare to Kling 4.0 Flash?

Kling 4.0 Flash is a streamlined companion mode built for rapid iteration and lower-cost generation, trading away maximum clip length, 1080p resolution, and keyframe controls. According to Kling's user guide FAQ and October 2026 trade coverage from ijiwei, here is how the two modes differ.

Kling 4.0 vs Kling 3.0: what changed?

Compared to Kling 3.0 (released February 5, 2026), Kling 4.0 doubles maximum single-take duration and expands native audio and multimodal controls. If you have used our Kling 3.0 AI video generator for 15-second clips with native audio, the comparison table and official upgrade list below show what changed between generations.

What Kling lists as new or upgraded in 4.0 (per the Kling 4.0 feature page FAQ):

• Flexible keyframes: Support for up to 10 keyframe images in a single generation.
• Multimodal references: Support for up to 15 combined reference assets (images, videos, elements, and voice profiles) with stronger character consistency.
• Motion and camera control: Improved fast motion, physical interactions, and multi-shot camera direction.
• Editing and recreation: Creative Recreation for swapping assets into reference video structures, plus targeted multi-clip video editing.

While evaluating generational upgrades, many creators also compare Kling 4.0 with ByteDance's Seedance 2.5 (launched July 31, 2026), which you can explore via the Seedance 2.5 AI video generator. Note that at the time of writing in early November 2026, Kling 4.0 does not yet have a published entry on independent blind-test leaderboards such as the Artificial Analysis Video Arena.

How much does Kling 4.0 cost?

As of November 2026, official Kling 4.0 API pricing starts at 1.0 unit (approximately $0.157) per second for 720p video and 1.5 units (approximately $0.235) per second for 1080p video when generating without an input video. When you supply an input video for reference or editing, the per-second unit rate drops, but the billed duration is calculated using a formula based on both your input and output clip lengths.

Using simple arithmetic on the official API rates above:

• A 10-second 1080p text-to-video clip (no input video) costs roughly 10 seconds multiplied by $0.235, or $2.35.
• A 30-second 1080p text-to-video clip costs roughly 30 seconds multiplied by $0.235, or $7.05.

For everyday creators using the consumer web app at kling.ai, Kling 4.0 has been accessible to all users on credit-based plans since October 15, 2026. Because we did not verify per-second consumer credit rates or subscription tier prices from an official consumer pricing table for this guide, you should check Kling's pricing page directly for current app plan details.

If you generate videos on Kling5 AI, credits scale with your selected resolution and clip duration; you can review plans and credit costs on the Kling5 AI pricing page.

What is still "coming soon" in Kling 4.0?

Although most of the Kling 4.0 feature set went live in October 2026, two announced capabilities are still labeled as "coming soon" in Kuaishou's official documentation as of November 2, 2026:

• 10-bit HDR video at 1080p and 4K: While 720p and standard 1080p are active now, 10-bit HDR color depth for 1080p and 4K outputs has not yet rolled out.
• 2-minute Video Extension: The ability to extend a generated video across multiple continuations up to 2 minutes in total duration is still pending release.

Separately, while 4K resolution is listed in the official documentation, the developer API pricing table does not list a 4K price yet (showing "-" in the rate column).

How to try Kling 4.0 today

You can try Kling 4.0 either through Kuaishou's official kling.ai website or on Kling5 AI, an independent AI video generator platform not affiliated with Kuaishou that provides direct browser access to Kling 4.0, Kling 3.0, and other leading video models.

To generate your first Kling 4.0 video on Kling5 AI, follow these five steps:

1. Open the workspace: Go to the Kling 4.0 AI video generator page on Kling5 AI.
2. Choose your generation mode: Select text-to-video, first-and-last-frame image-to-video, or reference-to-video (which supports up to 10 images and 5 reference videos).
3. Upload references and write your prompt: Add your visual references if applicable, then describe your scene's concrete visual details, camera movements (such as a slow push-in or tracking shot), and spoken dialogue. Add "no subtitles" or "no background music" if you want a clean visual or vocal track.
4. Set resolution, aspect ratio, and duration: Pick 720p or 1080p resolution, choose an aspect ratio from 16:9, 9:16, 1:1, or ultrawide 21:9, and set your clip length anywhere from 3 to 30 seconds.
5. Generate and download: Submit your task—your credit cost adjusts automatically based on duration and resolution—and preview and download your finished clip.

Sources

• Kuaishou Investor Relations: Kling AI Officially Launches Next-Generation Kling 4.0 (Oct 16, 2026)
• GlobeNewswire: Kling AI Launches Kling 4.0 Globally (Oct 16, 2026)
• Kling AI Blog: Introducing Kling 4.0 Early Access and Storytelling (Sep 30, 2026)
• Kling 4.0 Model User Guide (Oct 26, 2026)
• Kling 4.0 AI Video Generation Feature Page
• Kling Official API Changelog (Oct 28, 2026)
• Kling Official Developer API Video Pricing
• Kuaishou Investor Relations: Kling AI Launches 3.0 Model (Feb 5, 2026)