Kling 4.0 is official: 30-second video, 15 references and what it changes for AI video creators

What Kling 4.0 is and when it launched

Kling 4.0 is the new generation of the video model from Kling AI, the generative video platform of Kuaishou Technology. According to Kuaishou's official press release, issued from Hong Kong on October 16, 2026, the model officially launched on October 15, 2026. The same day Kling AI announced the global launch in a second press release from Beijing, which says Kling 4.0 is available now on the Kling AI website.

The launch caps a process that started in September: in the global release, Korean director Sung-Soo Kim describes using Kling 4.0 Preview during the September beta test. Alongside the full model there is Kling 4.0 Flash, a version built for faster, cheaper generation.

According to the company, Kling 4.0's three main promises are steadier motion even in dynamic scenes, more creative control and better narrative coherence over longer clips. Every figure in this article is claimed by Kuaishou and Kling AI, except where we cite The Decoder's hands-on comparison.

What's new: 30 seconds, 15 references and 10 keyframes

The most tangible change is length: Kling 4.0 generates video up to 30 seconds long in a single pass, with no need to stitch clips together. Here are the features announced at launch:

Read the number 15 carefully: it is a combined cap. The per-type maximums (10 images, 5 videos, 7 subjects) do not add up. References can also be storyboards, image grids, white-model videos (untextured blockouts) and depth-map videos, to guide performance, action, camera and pacing.

On audio, The Decoder reports two-channel stereo sound and improved lip sync in more than ten languages. Kling AI's release lists English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, with American, British and Indian English accents. If you work in another language, test it before you use it on client work.

2-minute extension, HDR and API: what is only announced

Some of the most talked-about features are not yet available to everyone. The launch releases use the future tense for these:

• Video extension up to 2 minutes, with multiple extensions of the same clip;
• 10-bit HDR output in 4K and 1080p;
• the 21:9 ultrawide cinema format;
• API access for enterprises and developers.

The sources do not agree on the 2-minute extension: Kling AI's global release says you can upload an existing video and extend it up to two minutes, while Kuaishou's release and The Decoder list it as "coming soon". The practical advice is simple: check your own account for the extension option before planning a project that depends on it.

Adobe Firefly is on the coming-soon list too: according to Kling AI's release, Kling 4.0 and Kling 4.0 Flash will be available "soon" inside Firefly. Neither the press releases nor The Decoder give prices or credit costs for Kling 4.0 or its API.

Kling 4.0 Flash and how it compares with Seedance and Wan

Kling 4.0 Flash is the lightweight variant: according to The Decoder it is faster and cheaper but tops out at 720p and 20 seconds. It suits drafts, prompt tests and high-volume content; for final delivery, the full model makes more sense.

How good is it really? The Decoder compared Kling 4.0 with Alibaba's Wan 3.0 and ByteDance's Seedance 2.5. Its verdict is cautious: Kling 4.0 holds its own but does not pull ahead, especially on shot logic, physics and natural dialogue. The author also suggests that progress in video models may be starting to slow down.

The company supplies its own figures: Kling AI says it has served more than 100 million users worldwide and more than 50,000 enterprise customers across advertising, film, TV and gaming, and Kuaishou notes that two works created with Kling AI won Gold Lions at Cannes Lions 2026. To decide which model fits your kind of video, our guide Kling vs Seedance vs Veo (in Italian) compares their strengths, and here is our Seedance guide.

What it means for you as an AI video creator

For creators, Kling 4.0 shifts the work from "generating clips" to directing a scene. Here is how to use it right away:

1. Build a reference kit. Before writing the prompt, gather images of the character, the product and the location, plus a short clip of the camera move you want. With 15 references you can keep the same face and the same product across the whole clip.
2. Think in keyframes. With 10 key images you can map the arc of a 30-second ad: opening, problem, product, result, close. It is the most reliable way to get a story instead of a random sequence.
3. Use Flash for tests. Try shots and prompts with Flash and switch to the full model only once the scene works: you save credits and time.
4. Recycle formats that work. With Creative Recreation you can take a video that already performed and swap the product or the character, which is handy for ad variants and UGC-style content.
5. Do not rely on announced features yet. Check the 2-minute extension, 4K HDR and API in your own account before promising them to a client.

If you are starting from scratch, our complete Kling AI guide (in Italian) covers how it works, what it costs and how Motion Control works. To write prompts that make the most of camera moves, see our guide to cinematic AI video prompts. It is the same race toward ever more capable models we saw with Meta Muse Wonder. And if you want to turn these tools into films, ads and cartoons with a complete method, the AI Filmaking program walks you through it step by step, from concept to storyboard to motion control. If you need a finished video without learning the workflow, our custom 30-second AI video service does it for you; for ad variants, 500 Creative Ads Prompts gives you angles, hooks and UGC scripts to test; here is also how we use artificial intelligence.