Kling AI has released version 4.0 of its video generator, extending single-pass clips to 30 seconds, adding two-channel stereo audio and lip sync in more than ten languages, and introducing a lighter Kling 4.0 Flash model. The update, reported on October 19 by The Decoder, is mostly about control: more reference material in, more consistent shots out. API pricing was not announced at launch.

What changed in Kling 4.0

The headline specification is length. Users can generate up to 30 seconds of video in one pass, and the system accepts up to ten keyframe images to pin down scene transitions, character states and story beats. Prompt comprehension has grown to 8,000 tokens, which leaves room for a detailed shot list rather than a single sentence.

The “Omni Reference” system, which feeds existing material into a generation, now takes up to 15 reference assets: up to 10 images, 5 videos with a combined maximum of 30 seconds, and 7 elements. Kling says storyboards, wireframes, image grids, white-model videos and depth-map videos all work as references. A new “Creative Recreation” feature treats an existing video as a template, so a product or character can be swapped while expressions, body movement, camera angles or backgrounds are adjusted without disturbing the rest of the clip, according to the company.

Sound and language

Kling 4.0 adds two-channel stereo audio and improved lip sync across more than ten languages, among them German, English, Japanese, Korean and Spanish, as well as various dialects and accents. In-video text generation now works in several languages and, Kling says, stays stable during camera movement. For creators who localize clips, that could reduce the work of re-recording or re-shooting for each market, although we have not tested how well the sync holds up in practice.

The Flash variant and agent access

Alongside the main model, Kling 4.0 Flash is designed for faster generation at lower cost. The trade-off is quality ceiling: it tops out at 720p and 20 seconds. The creation interface has also been redesigned, and Kling 4.0 can now be reached through MCP and a command-line interface, which makes it possible to wire video generation into automated workflows driven by AI agents. If you are new to the protocol, our guide to the Model Context Protocol explains how tools are exposed to agents.

Two features were listed as “coming soon”: 10-bit HDR output at 4K and 1080p, and a video extension function that would stretch clips to up to two minutes. API support and pricing are also still to come.

How it stacks up

The Decoder compared Kling 4.0 with Wan 3.0 and Seedance 2.5, both from Chinese competitors, in a video of its own. Its assessment was that Kling 4.0 holds its own but does not pull ahead, especially on logical shot sequencing, physics and natural-sounding dialogue. The outlet suggested this could be an early sign that the breakneck pace of progress in AI video is starting to level off, while noting that the next Wan and Seedance releases will tell more. That is one reviewer’s reading, not a benchmark result, and Kling’s own claim that the model should close the gap between AI-generated and professional video is a marketing statement.

Why it matters

Longer single-pass clips matter because stitching many short generations together is where consistency tends to break: faces drift, props change, lighting jumps. Keyframes and multi-asset references are the practical answer, since they let a producer fix the story beats and let the model fill the gaps. Native stereo audio removes another post-production step. For a deeper look at how generation, editing and audio fit together, see our AI video workflow for creators.

What to watch next

  • Kling 4.0 API pricing and availability, which determines whether studios and apps can build on it.
  • The HDR and two-minute extension features moving from “coming soon” to released.
  • Independent side-by-side tests against Wan and Seedance.
  • How Kling 4.0 Flash quality compares with the full model at 720p.
  • Licensing and disclosure rules for AI-generated video in advertising.

Key takeaways

  • Kling 4.0 generates up to 30 seconds per pass and accepts up to ten keyframes.
  • Omni Reference takes up to 15 assets, including images, videos and structural references.
  • Stereo audio and lip sync in more than ten languages are new.
  • The Flash variant is limited to 720p and 20 seconds.
  • API pricing, HDR output and two-minute extension are not yet available.

Frequently asked questions

How long can Kling 4.0 videos be?

Up to 30 seconds in a single pass for the main model, and up to 20 seconds for the Flash variant. A feature to extend clips to two minutes is listed as coming soon.

Does Kling 4.0 generate audio?

Yes. The release adds two-channel stereo audio and lip sync in more than ten languages, according to Kling AI.

Can developers use it in automated workflows?

Kling says version 4.0 can be accessed through MCP and a CLI. Standard API support is listed as coming soon, with pricing not yet announced.