Jiqizhixin (机器之心)→ original

Kling 4.0 Launches With Native 30-Second Video Generation and Combat Scenes

Kuaishou officially launched Kling 4.0, the new version of its video generator, on the evening of October 15. The headline features are native 30-second output in a single pass and a push into combat scenes, rather than only the dialogue and close-ups where Kling was strongest. In a closed test, users abroad made a fake trailer for a historical war drama called "Giant Commander": cavalry on a hillside, boulders hitting a fortress gate, a rain of arrows in a gorge. The original source does not name pricing, resolution or API access. • Kling 4.0 has been available since October 15. • Native 30 seconds, with no stitching of fragments. • The focus has shifted to action and crowd scenes.

AI-processed from Jiqizhixin (机器之心); edited by Hamidun News
Kling 4.0 Launches With Native 30-Second Video Generation and Combat Scenes
Source: Jiqizhixin (机器之心). Collage: Hamidun News.
◐ Listen to article
⊞Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceJiqizhixin (机器之心) ↗
Editor-in-Chief
Key metric #1Kling 4.0 has been available since October 15.
Key metric #2Native 30 seconds, with no stitching of fragments.
Key metric #3The focus has shifted to action and crowd scenes.

Kling 4.0 officially launched on the evening of October 15. The key number in the release is 30 seconds of video, which the model generates natively in a single pass, without manually stitching together short fragments. The second piece of news is that the developers are betting on action. Until now, many creators chose Kling for "文戏" (wenxi), scenes built on dialogue, facial expression and an actor's performance. Now the team is showing that the model can also handle "武戏" (wuxi), combat sequences.

What Testers Managed to Shoot in the Closed Test

Several weeks before the public launch, during internal testing, users abroad got access to the model and started assembling clips. The most prominent example is "巨将", which can be translated as "Giant Commander." It is a trailer for a historical war drama. Cavalry covers a hillside, heavy stones fly toward a city tower, and arrows rain down into a gorge. The source notes that the scale of the shot is impressive. The excerpt of the article available to us cuts off in the middle of a description of the camera work, so any assessment of the cinematography will have to wait for our own tests.

Scenes like this are hard for generative models. The frame holds many objects at once, each with its own trajectory, while the camera moves. If the model does not maintain physics and coherence well, the crowd turns into mush and projectiles behave arbitrarily.

Why 30 Seconds Matters

Most video generators produce clips of 5–10 seconds. To get a one-minute sequence, the creator stitches pieces together, and something usually breaks at the seams: a character's costume changes, the lighting shifts, the geography of the scene falls apart. In theory, native 30 seconds reduces the number of such seams. For a trailer or an ad spot, that is roughly ten shots that must hold a single visual style.

The limitation is just as obvious. The longer the continuous fragment, the higher the risk of accumulated artifacts toward the end. How well Kling 4.0 handles this will be shown by independent tests, not promo clips.

What Is Known About the Specs

  • Launch date: October 15, 2026, evening.
  • Duration: up to 30 seconds natively, a "direct output" with no stitching.
  • Release focus: combat and large-scale scenes, cinematic presentation.
  • Closed testing: ran before the release, and users abroad were already publishing clips.
  • Resolution, frame rate, audio, pricing and API access are not specified in the available excerpt of the material.

What to Test in Practice

If you make video for production, advertising or game trailers, start with three tests. First, a crowd scene with 30 seconds of continuous action. Watch how the last third of the clip behaves. Second, the same character from several angles within a single generation. Check whether the armor, face and weapon stay consistent. Third, interaction physics: hits, debris, smoke and crowd behavior.

At the same time, compare the cost per second of finished footage with what you spend now on stitching and regenerating short clips. That difference will show whether switching to the new version pays off. For now, it is better to rely on independent breakdowns and your own trial runs than on curated demos.

⧉ Story

❓ Key Questions Answered

What Testers Managed to Shoot in the Closed Test?

Several weeks before the public launch, during internal testing, users abroad got access to the model and started assembling clips. The most prominent example is "巨将", which can be translated as "Giant Commander." It is a trailer for a historical war drama. Cavalry covers a hillside, heavy stones fly toward a city tower, and arrows rain down into a gorge. The source notes that the scale of the shot is impressive. The excerpt of the article available to us cuts off in the middle of a description of the camera work, so any assessment of the cinematography will have to wait fo

Why 30 Seconds Matters?

Most video generators produce clips of 5–10 seconds. To get a one-minute sequence, the creator stitches pieces together, and something usually breaks at the seams: a character's costume changes, the lighting shifts, the geography of the scene falls apart. In theory, native 30 seconds reduces the number of such seams. For a trailer or an ad spot, that is roughly ten shots that must hold a single visual style. The limitation is just as obvious. The longer the continuous fragment, the higher the risk of accumulated artifacts toward the end. How well Kling 4.0 handles this will

What Is Known About the Specs?

Launch date: October 15, 2026, evening. Duration: up to 30 seconds natively, a "direct output" with no stitching. Release focus: combat and large-scale scenes, cinematic presentation. Closed testing: ran before the release, and users abroad were already publishing clips. Resolution, frame rate, audio, pricing and API access are not specified in the available excerpt of the material.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace

Access Frontier LLMs & Multimodal AI in One Workspace

Discover NomadNet AI (nomadnet.ai) — unified access to frontier language models, research synthesis, and multimodal workflows.

Try NomadNet AI →Launch NomadNet Agent
What do you think?
Loading comments…