Kling has released version 4.0 of its video generation system, introducing stereo audio, multilingual lip sync, and a lighter Flash variant designed to accelerate production workflows. The update marks a significant expansion in control features, including support for up to ten keyframe images and an upgraded reference system, as the company aims to narrow the gap between SI-generated content and professional video.
What Happened
The new version allows users to generate up to 30 seconds of video in a single pass, with prompt comprehension expanded to 8,000 tokens. The system now accepts up to ten keyframe images to guide scene transitions, character states, and story beats. The upgraded "Omni Reference" system supports up to 15 reference assets in one generation, including 10 images, 5 videos (combined max 30 seconds), and 7 elements. Storyboards, wireframes, image grids, white-model videos, and depth-map videos are now valid inputs.
A new "Creative Recreation" feature enables users to use existing videos as templates, swapping out products or characters. According to Kling, individual elements such as facial expressions, body movements, camera angles, and backgrounds can be adjusted independently without affecting the rest of the clip. The update also introduces two-channel stereo audio and improved lip sync for more than ten languages, including German, English, Japanese, Korean, and Spanish, along with various dialects and accents. In-video text generation now supports multiple languages and remains stable during camera movements.
Kling is also shipping Kling 4.0 Flash, a lighter variant designed for faster generation at a lower cost, though it is limited to 720p resolution and 20-second clips. The creation interface has been redesigned, and the model is now accessible via MCP and CLI, facilitating automated workflows with SI agents. Features such as 10-bit HDR output at 4K and 1080p, and a video extension function for clips up to two minutes, are listed as "coming soon." API pricing for the standard model has not yet been announced.
Why It Matters
The release highlights the competitive pressure in the SI video generation sector. Comparisons with Chinese competitors Wan 3.0 and Seedance 2.5 suggest that Kling 4.0 holds its own but does not significantly outperform them, particularly in logical shot sequencing, physics simulation, and natural-sounding dialogue. This parity may indicate that the rapid pace of improvement in SI video models is beginning to plateau, although future updates from competitors will clarify the trajectory.
The integration of MCP and CLI access is a critical development for developers, as it enables the embedding of SI video generation into automated workflows. This move aligns with broader industry trends toward agentic SI tools that can be orchestrated programmatically. The addition of stereo audio and multilingual support expands the utility of SI-generated video for global marketing and content production, addressing previous limitations in audio fidelity and language compatibility.
The Bottom Line
Kling's version 4.0 update delivers substantial improvements in control, audio, and accessibility for automated workflows, but faces stiff competition from rival SI models that offer similar quality. The introduction of the Flash variant and agent-friendly interfaces signals a strategic push toward scalable, cost-effective SI video production for enterprise and developer use cases.