Kling 4.0 Is Coming: What We Know, and How to Prepare on WaveSpeedAI

Kling 4.0 rumors now point to two video variants, Preview and Flash, alongside a new Kling Image 4.0 image generation model. The reported upgrades include approximately 3–30 seconds per standard video generation, an experimental long-video mode reaching about 120 seconds, up to roughly 10 keyframes, and support for more reference material.

Updated September 28, 2026: These are reported preview features, not confirmed launch specifications. The earlier enterprise-testing timeline discussed in this article was also unverified and should not be read as a release announcement. Public availability, API access, pricing, and final limits still need confirmation.

What has been reported about Kling 4.0

The latest claims concern leaked development-mode settings. ArtRealmAI’s September 24 report attributes the leak to a post by Kol Tregaskes. These are reports about a possible upcoming product, not documentation from Kuaishou.

Area	Reported Kling 4.0 update	How to read it
Video variants	Kling 4.0 Preview and Kling 4.0 Flash Preview	Two reported names; performance and pricing differences are unverified
Single-generation duration	Approximately 3–30 seconds for Preview	Do not assume Flash has the same upper limit
Experimental long video	Up to approximately 120 seconds	A separate experimental mode, not a promised standard clip length
Keyframes	Up to roughly 10	Reported control points, not a confirmed number of shots
Reference material	More reference inputs	Final counts, supported types, and combinations remain unconfirmed
Image generation	Kling Image 4.0 reportedly on the way	A separate image model; release timing and specifications are unconfirmed

Third-party coverage of the preview settings also distinguishes Preview, Flash, keyframe control, and experimental long video. Multiple sites repeating a leak do not independently verify it.

Preview and Flash: keep the limits separate

The two reported video variants could give creators different options, but the name Flash alone does not establish faster generation, lower prices, or a particular quality tradeoff. Until there are official specifications and comparable outputs, neither variant should be assigned a measured advantage.

From 15 seconds to a reported 30 seconds

Kling’s official VIDEO 3.0 user guide documents flexible 3–15 second generation. The rumored Preview range of 3–30 seconds would double that upper limit if it ships.

For a creator, that could mean fitting a longer product reveal or a complete action into one generation. It would not, by itself, prove better motion or consistency: a longer timeline also gives visual errors more time to accumulate.

The approximately 120-second long-video claim is experimental. It should not be treated as evidence that ordinary Preview or Flash requests can deliver a continuous two-minute shot, or that all reference and audio controls work in that mode.

More keyframes and reference material

The reported up-to-10-keyframe workflow could offer more checkpoints for directing how a scene develops. Keyframes and references serve different purposes: keyframes guide moments along a timeline, while references describe subjects, environments, or other visual details. Ten keyframes should not be advertised as ten separate shots.

Expanded reference support could make room for more character, product, and location material. Exact reference budgets and which input types can be combined still need confirmation, especially across Preview, Flash, and experimental long video.

Kling Image 4.0 is also reportedly coming

The same rumors name Kling Image 4.0, a forthcoming image generation model. If released, it could be useful for creating still assets to use in a video workflow, but that is a possible creative use, not a confirmed integration. Image-model rumors do not establish video resolution, and we do not have verified Image 4.0 pricing, API parameters, or a release date to publish.

Why this matters for the market

The framing in the community posts is telling: Kling 4.0 is positioned as the model that breaks the current dominance of Seedance 2.0 and 2.5 in commercial video work, particularly for short drama, e-commerce, and advertising.

That framing matches what we see in usage. Seedance 2.5 has become a default choice for many production teams because of its motion quality and native audio. Kling 3.0 and O3 already compete on 4K output, multi-shot storytelling, and video editing. A Kling 4.0 that closes the gap on character consistency and audio sync would give commercial buyers two credible top-tier options, and price competition between them tends to benefit everyone building on top.

For an API user, the practical takeaway is that you should not be locked to one vendor’s model family. WaveSpeedAI exposes Kling, Seedance, Veo, WAN, MiniMax, and others behind one API, so switching or A/B testing is a matter of changing the model path in the request URL.

The Kling models live on WaveSpeedAI today

While we await confirmed 4.0 details, the current Kling lineup on WaveSpeedAI is the broadest it has ever been. Every model below is available now, billed per second of generated video, with no subscription. Current rates are on the Kling 3.0 API page and the Kling O3 API page.

Kling 3.0: the current flagship

The Kling 3.0 family covers text-to-video, image-to-video, and motion control at three quality tiers.

Model	Task	Model path
Kling 3.0 Standard	text-to-video, image-to-video	kwaivgi/kling-v3.0-std/text-to-video, kwaivgi/kling-v3.0-std/image-to-video
Kling 3.0 Pro	text-to-video, image-to-video	kwaivgi/kling-v3.0-pro/text-to-video, kwaivgi/kling-v3.0-pro/image-to-video
Kling 3.0 4K	text-to-video, image-to-video	kwaivgi/kling-v3.0-4k/text-to-video, kwaivgi/kling-v3.0-4k/image-to-video
Kling 3.0 Standard Motion Control	video-to-video	kwaivgi/kling-v3.0-std/motion-control
Kling 3.0 Pro Motion Control	video-to-video	kwaivgi/kling-v3.0-pro/motion-control

Kling 3.0 introduced native 4K output, multi-shot prompts, and improved physical realism. Motion control transfers the movement from a reference clip onto a still image, which is the fastest way to get a specific dance, gesture, or action without prompting for it.

Kling 3.0 Turbo: for iteration

The Turbo variants trade a little fidelity for speed and a lower price, and they support the same single-prompt and multi-shot inputs as the full 3.0 models.

Model	Model path
Kling V3 Turbo Standard	kwaivgi/kling-v3-turbo-std/text-to-video, kwaivgi/kling-v3-turbo-std/image-to-video
Kling V3 Turbo Pro	kwaivgi/kling-v3-turbo-pro/text-to-video, kwaivgi/kling-v3-turbo-pro/image-to-video

Turbo is the right place to develop prompts. Once a prompt and reference set work on Turbo, they transfer to the Standard, Pro, or 4K endpoints unchanged.

Kling O3: reference-to-video and video editing

Kling O3, also marketed as Kling 3.0 Omni, is the unified multimodal branch of the family. Beyond text and image input, it accepts up to seven reference images, a reference video, and reusable Kling Elements for consistent characters across clips. It also edits existing footage.

Tier	Tasks	Model path prefix
O3 Standard	text, image, and reference to video; video edit	kwaivgi/kling-video-o3-std/
O3 Pro	text, image, and reference to video; video edit	kwaivgi/kling-video-o3-pro/
O3 4K	text, image, and reference to video; video edit; video reference	kwaivgi/kling-video-o3-4k/

The 4K tier also has dedicated video-edit and video-reference endpoints, covered in detail in our Kling O3 4K video edit and video reference post.

If the Kling 4.0 reports about character consistency are accurate, O3 with Elements is the closest thing available today, and the workflows you build around it will carry over.

Kling 2.6 and earlier

Kling 2.6 Standard and Pro remain available for teams that have tuned prompts against them, along with Kling 2.5 Turbo, Kling LipSync, Kling AI Avatar, and the Kling audio models. See the full list on the Kling 3.0 API page.

Try Kling without code: the WaveSpeedAI Video Generator

Not every test needs an API key. The WaveSpeedAI Video Generator runs Kling 3.0 in the browser next to Seedance 2.5, Google Veo 3.1, WAN 3.0, MiniMax, Vidu Q3, Grok Imagine, and more than a dozen other models.

What it does:

• Text-to-video and image-to-video in one interface, with the same prompt reusable across models.
• Side-by-side model comparison, which is the fastest way to decide whether Kling, Seedance, or Veo fits a given brief before committing API budget.
• Resolution up to 4K, durations from 3 to 15 seconds, and native audio on the models that support it.
• No watermark on output, and a free tier to start.

Watch this blog for confirmed Kling 4.0 availability on WaveSpeedAI, including which variants and workflows are supported.

How to prepare your pipeline for Kling 4.0

Every Kling model on WaveSpeedAI uses the same request shape: submit a JSON body to the model path, receive a prediction ID, poll for the result. You can reuse that integration pattern, but check the published 4.0 schema before switching: durations, reference inputs, and other parameters may differ.

import os
import time
import requests

API_KEY = os.environ["WAVESPEED_API_KEY"]
BASE = "https://api.wavespeed.ai/api/v3"

# Check the published model schema before switching to Kling 4.0.
MODEL = "kwaivgi/kling-v3.0-pro/text-to-video"

resp = requests.post(
    f"{BASE}/{MODEL}",
    headers={"Authorization": f"Bearer {API_KEY}", "Content-Type": "application/json"},
    json={
        "prompt": "A tailor pins fabric on a mannequin in a sunlit studio, slow dolly-in, dust in the light beams",
        "duration": 5,
        "aspect_ratio": "16:9",
    },
)
resp.raise_for_status()
prediction_id = resp.json()["data"]["id"]

while True:
    result = requests.get(
        f"{BASE}/predictions/{prediction_id}/result",
        headers={"Authorization": f"Bearer {API_KEY}"},
    ).json()["data"]
    if result["status"] == "completed":
        print(result["outputs"][0])
        break
    if result["status"] == "failed":
        raise RuntimeError(result.get("error"))
    time.sleep(2)

A few things worth doing now:

1. Keep the model path in configuration, not in code. Teams that did this for the 2.6 to 3.0 transition upgraded in an afternoon.
2. Build a fixed prompt set of 10 to 20 briefs that represent your real workload, and record the outputs from Kling 3.0 and Seedance 2.5 today. When 4.0 arrives, run the same set and compare consistency, motion, and cost per approved clip rather than reacting to demo reels.
3. Prototype on Turbo, finalize on Pro or 4K. Kling V3 Turbo is the lowest-cost place in the Kling 3.0 family to iterate. The final render on Pro or 4K is a path change.
4. Adopt Kling Elements if character consistency matters to you. Elements are reusable across O3 endpoints and are the most likely mechanism to carry forward into 4.0.

What happens next

We will publish an update on this blog as soon as Kuaishou confirms Kling 4.0 details, and again when the model is live on WaveSpeedAI. Until then, the Kling 3.0 API page is the place to check for the current model list and prices, and the Video Generator is the fastest way to see what the current generation can already do.

Related reading

• Kling 3.0 Omni Explained
• Kling O3 4K Adds Video Edit and Video Reference
• Seedance 2.5 vs Kling: Video Generation Compared
• What to Expect from Kling 3.0: A Technical Preview, our preview from before the 3.0 launch, for a sense of how these predictions have tracked

FAQ

Has Kuaishou officially announced Kling 4.0?

The reports discussed in this September 28, 2026 update are leaks, not an official specification. Preview, Flash, longer clips, additional keyframes, expanded references, and Image 4.0 remain unconfirmed; no verified public release date is provided here.

Will Kling 4.0 be available on WaveSpeedAI?

WaveSpeedAI has carried every major Kling release since V1.6, including Kling 3.0, Kling 3.0 Turbo, Kling 3.0 4K, and the Kling O3 family. We plan to make Kling 4.0 available as soon as API access opens, and will announce it on this blog and on the Kling 3.0 API page.

Which Kling model should I use today while waiting for 4.0?

Use Kling 3.0 Turbo for fast, lower-cost iteration, Kling 3.0 Standard or Pro for final 1080p clips, Kling 3.0 4K or Kling O3 4K for 4K delivery, and Kling O3 for reference-to-video and video editing. All of them share the same request and polling pattern, but adopting 4.0 will require checking its published input schema and limits.

Can I try Kling models without writing code?

Yes. The WaveSpeedAI Video Generator at wavespeed.ai/video-generator lets you run Kling 3.0 alongside Seedance 2.5, Veo 3.1, WAN 3.0, and other models from a browser, with text-to-video and image-to-video modes and no watermark.