Flux.art

Kling 4.0

Kuaishou Kling AI's next video model—longer takes, keyframe direction, and richer multimodal reference.
Coming soon on Flux.art. Full public release targeted for October 2026. Generate with Kling 3.0 today.

Coming SoonTry Kling 3.0

What is Kling 4.0?

Kling 4.0 is the September 2026 generation of Kuaishou's Kling AI video stack, announced on 28 September as the successor to Kling 3.0. Official materials and contemporaneous reporting describe a model built for longer, more directed takes rather than another five-second wallpaper loop. A single native generation can run from three to thirty seconds. Up to ten stills can sit on the timeline as keyframes so the clip unfolds between planned beats instead of inventing the whole arc from one sentence. Omni Reference can pull as many as fifteen multimodal inputs—images, video clips, a voice reference, and saved subjects—into one job, with published caps of ten images, five reference videos (thirty seconds combined), and seven elements. Prompts stretch to eight thousand tokens when you need a shot list, not a slogan. Picture output is listed at 720p, 1080p, and 4K, with a new 21:9 ultrawide frame alongside 16:9, 1:1, 9:16, and Auto. Audio moves to two-channel stereo with tighter lip-sync, and spoken coverage expands to include German and French on top of Chinese, English, Japanese, Korean, Spanish, and Portuguese, plus regional Chinese dialects and English accents. On-screen type is called out as a 4.0 addition: Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and emoji that are meant to stay readable when the camera moves. Two items are still labeled coming soon on Kling's own spec: 10-bit HDR at 1080p and 4K, and multi-pass extension that can grow a clip toward two minutes. The full public model is targeted for October 2026 without a named day. A lighter sibling, Kling 4.0 Flash, opened limited early access on 28 September to Kling Ultra Yearly (Black Gold) members: three to twenty seconds at 720p SDR, first-frame image-to-video, and Omni Reference, but not the 4K or HDR path. On Flux.art, Kling 4.0 is coming soon. We will add text-to-video and image-to-video when the full model is available to our stack. Until then, Kling 3.0 already runs here for multi-shot three-to-fifteen-second clips at 720p, 1080p, or 4K with native audio. This page is a capabilities brief so you can assemble briefs, keyframes, and reference packs before the switch turns on—not a generate button dressed as a launch.

3–30 Seconds in One Pass

Double the native ceiling versus Kling 3.0's 15-second cap. Write a scene that needs a hold, a turn, and a closer without stitching three mismatched renders.

Up to 10 Timeline Keyframes

Place stills as beats. Kling 4.0 interpolates continuous video between them so character state and scene changes follow the order you planned.

Omni Reference, 15 Inputs

Combine up to ten images, five clips, a voice, and seven saved subjects in one request so faces, products, and motion style stay on brief.

Coming Soon on Flux.art

Kling 4.0 is not live on Flux.art yet. Watch this page for launch. Until then, create with Kling 3.0 in the video generator.

Why Kling 4.0 Matters for Directed Video

Kling 4.0 is aimed at teams who already outgrew a single locked-off take: longer scenes, a keyframe storyboard, and references that behave like a production bible—not a moodboard the model politely ignores.

Most consumer AI video still lives in the five-to-ten-second band. That is enough for a product spin or a hook, and it is exactly where Kling 3.0 on Flux.art already works well with a three-to-fifteen-second ladder. Kling 4.0's headline change is duration as a first-class production unit: three to thirty continuous seconds in one pass. Official notes pitch the extra time as room for an unbroken take—an establishing move, a line of dialogue, a reaction, and a button—without cutting to a second generator and hoping identity survives the join. You can still work short. The floor remains three seconds, so a social hook does not have to inflate to a half-minute. The useful shift is that a thirty-second brand spot or a one-scene short-drama beat can be one job instead of a timeline of stitches. Kling also lists multi-pass continuation that can extend footage toward two minutes; that path is still marked coming soon, so plan first releases around the native thirty-second cap. On Flux.art we will expose duration once the model is live. Until then, treat this as a planning number: write the scene to thirty seconds, then cut a fifteen-second Kling 3.0 proof if you need pixels this week.

Kling 4.0 Capabilities at a Glance

What Kling announced for the full October model versus Flash—and what Flux.art will support only after launch.

Text-to-Video and Image-to-Video

Full 4.0 lists both. Flash image-to-video is first-frame only. Flux.art will enable the modes we actually wire, marked coming soon until then.

Multi-Keyframe Storyboarding

Up to ten stills on a timeline. The model generates the in-betweens so scene and character state follow your beat sheet.

Omni Reference and Video Edit

Up to fifteen mixed inputs, plus an edit path of one primary clip and four supports inside a thirty-second budget.

Stereo Audio and On-Screen Type

Two-channel sound, tighter lip-sync, wider language and dialect coverage, and readable supers that are meant to survive camera moves.

3–30 Seconds, 720p to 4K, 21:9

Flexible duration. Full model lists 720p, 1080p, and 4K. Flash is 720p SDR. Ultrawide 21:9 is new. HDR remains coming soon.

Coming Soon on Flux.art

No generate button yet. Use Kling 3.0 for production today. This page will update when Kling 4.0 is available in our video generator.

Where Kling 4.0 Is Aimed

Use cases that benefit from a thirty-second scene, a keyframe pack, and references that keep talent and product stable.

Ecommerce and ad video planned for Kling 4.0

Ads, Ecommerce, and Product Spots

A thirty-second hero can open wide, cut to a detail still you pinned as a keyframe, and close on packaging with readable type. Keep the bottle and face locked with Omni Reference instead of hoping a long prompt remembers the SKU.

Cinematic 21:9 storytelling with Kling 4.0

Cinematic and Ultrawide Story Beats

21:9 framing plus a longer take is built for trailers, short-drama scenes, and documentary-style walk-and-talks. Keyframes can mark a location change without starting a second job.

Localized social video with Kling 4.0

Localized Dialogue and Social Cuts

Stereo lines in more languages and dialects, plus on-screen type and emoji for 9:16. Plan a Flash-speed draft at 720p when it lands, then step to full 4.0 for the finishing resolution.

How Kling 4.0 Will Work on Flux.art

The model is coming soon. Use this path to prepare—and keep shipping on Kling 3.0 until launch:

1

Prepare the Brief and Keyframe Pack

Write the scene as a beat sheet: who is in frame, what happens, where the camera sits. Collect up to ten stills for key moments and product plates so you are ready when reference slots open.

2

Watch This Page for Launch

Kling 4.0 is not in the Flux.art video generator yet. When we enable it, this model page and the generator will show the live duration, ratio, and sound controls the API exposes.

3

Create Today with Kling 3.0

Need pixels this week? Run Kling 3.0 for text-to-video or image-to-video, 3–15 seconds, 16:9 / 9:16 / 1:1, optional sound, and 720p to 4K.

4

Move the Same Brief to 4.0 Later

Keep prompts, stills, and brand rules in one folder. After launch, drop the same pack into Kling 4.0 for the longer take, extra keyframes, and Omni Reference.

Frequently Asked Questions

Common questions about Kling 4.0 and its coming-soon status on Flux.art.

Kling 4.0 Is Coming Soon on Flux.art

We will enable Kling 4.0 when the full model is available. Until then, direct multi-shot scenes with Kling 3.0—native audio, locked subjects, and up to 4K.

Coming SoonTry Kling 3.0