![What is Kling 4.0? Everything You Need to Know [2026]](https://refinery.fal.media/url/https%3A%2F%2Fv3b.fal.media%2Ffiles%2Fb%2F0aac880f%2FiTfQ74jiVsnY8JWOk7yeH.jpg/tr:w-1920,q-80/iTfQ74jiVsnY8JWOk7yeH.webp)
Kling 4.0 is on its way, and we can't wait to put it in your hands on fal.
The specs and a big batch of sample clips are already out, so this guide covers what the model can do before its big launch this year.
TL;DR
Kuaishou's Kling AI has set October 2026 for the full launch of Kling 4.0, its next video generation model.
Single generations run for up to 30 seconds and can be steered with as many as 10 keyframe images.
A faster version called Kling 4.0 Flash has been in limited early access since September 28.
Omni Reference can pull up to 15 references into one request. Existing footage can go in too, up to 5 clips per edit.
Output goes as high as 4K with stereo sound, while 10-bit HDR output is still on its way.
Kling 4.0 is coming to fal soon, and we're confident fal will be the best place to run it. That's because you'll test prompts in our playground and ship the same request through the API that already runs Kling O3 and Kling 3.0.
What is Kling 4.0?
Kling 4.0 is the newest generation of Kuaishou's Kling video models.
The full Kling 4.0 lands in October, but early access to the faster Kling 4.0 Flash opened in late September.
The full model's longest single clip runs 30 seconds, pitched by Kling as room for unbroken takes.
Kling has already shown the model in dozens of sample clips.
When can you use Kling 4.0?
October is the month Kling has given for the full release, without a specific day attached so far.
Flash went out first, on September 28, to a limited group of Kling users through early access.
We're bringing Kling 4.0 to fal soon and will post the endpoints and settings on its model page when it goes live.
💡 Note: I've flagged the couple of announced features Kling still labels as coming soon wherever they come up below. Every spec and capability claim in this guide comes from Kling's release notes and showcase clips.
What's new in Kling 4.0?
Eight rows in Kling's 4.0 spec table carry a new tag, marked in our version below.
| Spec | Kling 4.0 |
|---|---|
| Text-to-video | ✅ Yes |
| Image-to-video | ✅ Yes |
| First and last frames | ✅ Yes |
| Multi-keyframes | ✅ Up to 10 keyframe images (new) |
| Omni Reference | ✅ Yes |
| Video duration | 3 to 30 seconds (new) |
| Resolution | 720p, 1080p, 4K |
| Dynamic range | 10-bit HDR at 1080p and 4K (new, coming soon) |
| Aspect ratio | 21:9 (new), 16:9, 1:1, 9:16, Auto |
| Audio output | Two-channel stereo (new) |
| Languages | Chinese, English, Japanese, Korean, Spanish, Portuguese, German (new), French (new), and more |
| Chinese accents and dialects | Beijing, Taiwanese Mandarin, Northeastern, Sichuan dialect, Cantonese, and more |
| English accents | American, British, Indian, and more |
| On-screen text | Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and more, plus emoji (new) |
| Text input length | Up to 8,000 tokens (new) |
| Reference images | Up to 10 |
| Reference videos | Up to 5, with 30 seconds combined |
| Reference audio | Voice reference only |
| Reference elements | Up to 7, including up to 3 video-based |
| Total references | Up to 15 across images, videos, voice references, and elements |
| Video editing | Up to 5 input clips including 1 primary video, each 3 to 30 seconds, 30 seconds combined |
How much of a story fits into one Kling 4.0 generation?
A single Kling 4.0 generation can be anything from 3 to 30 seconds long.
Kling says a clip with as many as 10 keyframe images will unfold between them the way you planned.
Kling also claims each keyframe can pin down a character's state or a change of scene along the way.
Prompts can now run to 8,000 tokens, though the more interesting claim concerns very short ones.
The model is meant to build a thin prompt out the way a director would.
Kling's own example is the six-word prompt "An old man ranting about AI."
Kling's notes promise more natural camera work across push-ins, pans, orbits and shot-size changes.
Video extension will give longer pieces another route by letting you upload a clip and extend it multiple times.
Kling caps extended videos at 2 minutes and still lists the feature as coming soon.
Which inputs does Omni Reference accept?
Omni Reference is how existing material gets into a Kling 4.0 generation.
Picture a two-person kitchen ad built from four stills, two of the actors and two of the product.
That's four of the 10 image slots, and a 12-second motion reference would leave 18 seconds of the video allowance.
The spec table above has the rest of the caps, up to 15 references overall.
On the image side, Kling accepts three-view sheets, image grids, wireframes and keyframes alongside ordinary subject photos.
Clips steer the performance and the camera, including depth-map and white-model videos.
Kling says those give you tighter control over motion and timing.
The AI lab claims that references combine more reliably in 4.0 and keep characters closer to the brief.
falMODEL APIs
The fastest, cheapest and most reliable way to run genAI models. 1 API, 100s of models
Can Kling 4.0 edit a video you already have?
A clip you already have can go back into Kling 4.0 for targeted changes.
Kling names expressions, body movement, camera position and overall style as things you can change, with the rest of the shot held steady.
Background swaps are on the list as well, along with replacing a character or removing an element from the scene.
An edit works from one primary video plus up to 4 supporting clips, all of them 3 to 30 seconds long and 30 seconds in total.
Creative recreation works a lot like editing, borrowing the pacing and composition of an existing video for new content.
A short-drama series could keep the shot rhythm of an earlier episode and swap in a different lead character.
Will Kling 4.0 footage look and sound more realistic?
Kling put realism at the top of its 4.0 highlights for both picture and sound.
Fast action and big camera moves are meant to stay smooth and stable, according to Kling, even through tracking and orbit shots.
Faces should carry more expression as well, down to a single glance.
On the audio side, two-channel stereo output is one of the rows Kling tags as new.
Kling also promises more accurate lip sync, and dialogue scenes will be the real test of that claim.
HDR is still on the way, and Kling's 10-bit mode will cover 1080p and 4K output once it ships.
In that mode, Kling expects shadows and highlights to hold more detail and night shots to show less color banding.
Ultrawide 21:9 framing is part of the supported spec for anyone composing wide cinematic shots.
Which languages and visual styles does Kling 4.0 cover?
Kling demonstrates 4.0 in vlog and documentary looks as well as claymation, felt textures, glitch art and 3D cyberpunk.
Its genre examples include game trailers and UGC-style product promos, among others.
Kling's dialogue languages include English, Chinese, Japanese, Korean, Spanish, Portuguese, German and French, among others.
German and French both carry new tags on Kling's spec table.
Accent support reaches Cantonese and Sichuan dialect on the Chinese side, along with regional English accents like Indian English.
On-screen text is new in 4.0, covering the same core languages plus emoji and logos.
Kling says the words stay clear and stable even when the camera moves or the scene changes.
How does Kling 4.0 Flash fit into the lineup?
Kling is selling Flash on speed and value to people who generate often.
The table below lines Flash up against the full model on the rows Kling published for both.
| Kling 4.0 | Kling 4.0 Flash | |
|---|---|---|
| Text-to-video | ✅ Yes | ✅ Yes |
| Image-to-video | ✅ Yes | ✅ First frame |
| Omni Reference | ✅ Yes | ✅ Yes |
| Video duration | 3 to 30 seconds | 3 to 20 seconds |
| Resolution | 720p, 1080p, 4K | 720p |
| Dynamic range | 10-bit HDR at 1080p and 4K (coming soon) | 8-bit SDR |
| Availability | October launch | Limited early access since September 28 |
Where will you be able to run Kling 4.0?
The best place to run Kling 4.0 will be on fal, which is where you can already access Kling O3, Kling 3.0, and 1,000+ other AI models behind one API.
You'll be able to try prompts in our playground first, then move the same request into production through the API, or use our fal Agent and MCP to generate media with LLMs like ChatGPT and Claude.
Endpoint names and pricing will go up on the Kling 4.0 model page at launch.
Editor's opinion: which Kling 4.0 limit should you plan around?
I'd plan a production around the 30-second cap on video input before I'd worry about output length.
Every reference clip counts against that cap, and so do subjects built from video.
Edits have a matching 30-second ceiling, since the primary clip and any supporting clips can't total more than that.
With a 3-second minimum per clip, one supporting clip caps the primary video at 27 seconds on Kling's published limits.
I'd expect a good chunk of early prep time to go into cutting source footage down before it ever reaches the model.
For teams with existing footage and voice talent, being able to feed both into one generation is exciting.
I'll want to see how the model copes with a long edit against that budget before building any pipeline on it.
That's the first test on my list once Kling 4.0 lands on fal.
Recently Added
Ready to run Kling 4.0 on fal?
When Kling 4.0 goes live, you'll reach it through the same API you already use for Kling 3.0 and the rest of our video models.
If you haven't made a fal account yet, you can set one up today and top up from $1 ahead of launch.
Frequently asked questions about Kling 4.0
Who makes Kling 4.0?
Kuaishou develops Kling 4.0 and releases it through its Kling AI platform. You can run its predecessor, Kling 3.0, on fal today.
Is there an exact Kling 4.0 release date?
Kling has committed to October 2026 for the full model but hasn't named a day. The release note doesn't say whether Flash access widens before then.
What is the maximum Kling 4.0 clip length?
A single generation tops out at 30 seconds, while Flash clips stop at 20.
The minimum for both models is 3 seconds, according to the spec tables Kling published for each of them.
What resolution does Kling 4.0 output?
Kling lists outputs up to 4K for the full model and 720p for Flash.
Frame shapes now include 21:9 ultrawide as a new option.
How many references can a Kling 4.0 generation take?
Up to 15, with images capped at 10 and video references capped at 5.
Up to 7 subjects fit in one request, and a subject can carry a specified voice.
Can Kling 4.0 speak with regional accents?
Yes, Kling lists several Chinese accents and dialects, including Cantonese, along with a range of regional English accents.
The spoken language list covers at least eight languages, English and Chinese among them.
Can you use Kling 4.0 Flash today?
Only members of Kling's limited early-access group can use it right now.
Kling built Flash for fast, frequent generation, but its spec table still lists Omni Reference support.
Does Kling 4.0 produce sound?
It does, with two-channel stereo audio and what Kling describes as more accurate lip sync.
Kling lists voice as the only type of audio reference, framed as a voice you give one of your subjects.
What are the limits on a Kling 4.0 edit?
Five input clips is the ceiling, one of them the primary video.
Each clip needs 3 to 30 seconds of runtime inside a 30-second cap for the whole set.
![10 Best Video-to-Video APIs in 2026 [Reviewed]](https://refinery.fal.media/url/https%3A%2F%2Fv3b.fal.media%2Ffiles%2Fb%2F0aa0ad7c%2FCnYBY_0niENSKUXkMaEIf_best-video-to-video-apis-2026.jpg/tr:w-1280,q-80/CnYBY_0niENSKUXkMaEIf_best-video-to-video-apis-2026.webp)
