Bottom Line Up Front (BLUF): Kling 4.0 generates continuous video clips from 3 to 30 seconds in a single pass. The model supports Omni Reference with up to 15 multimodal references per request (up to 10 images, 5 videos, 7 characters, and 3 voices), keyframe control, native realistic audio, and synchronized lip-sync. Resolutions include 720p and 1080p, with aspect ratios auto, 16:9, 9:16, and 1:1.
Kling 4.0 enables creating continuous videos up to 30 seconds long without splitting the prompt into short segments or stitching clips together. This delivers fluid motion, stable composition, and consistent storytelling throughout the entire scene.
Integrated audio synthesis automatically creates realistic environmental sound effects, foley, and accurate lip-sync for speaking characters. When an input video is provided, the original sound can be preserved.
Kling 4.0 accepts text prompts alongside multiple images, reference videos, character anchors, and voices. Each request supports up to 10 images and 5 video clips (up to 15 references in total).
In your scene description, reference uploaded assets with tags such as @image_1, @video_1, or character names. The model replicates the camera motion and pacing of the reference video while swapping subjects and backgrounds.
The keyframe mode anchors the opening and closing compositions, with up to four intermediate milestones. This gives directors precise control over pacing and story progression.
The model smoothly interpolates motion between designated frames, maintaining realistic physics, consistent lighting, and character fidelity across all shots.
Available durations range from 3 to 30 seconds at 720p or 1080p, with aspect ratios auto, 16:9, 9:16, and 1:1. The cost is quoted clearly before generation: a standard per-second rate applies for text/image inputs, and a lower per-second rate applies when reference videos are provided.
Tokens are deducted upon task creation and automatically refunded if generation fails or is rejected.
Kling 4.0 is available in the public API at POST /v1/videos/generations with model kling-4-0, and through the generate_video tool in the MCP server. All parameters are supported: duration, resolution, aspect_ratio, image_urls, video_urls, keep_original_sound, and keyframes.
Media assets can be provided as HTTP URLs or Base64 strings. Check task progress using the task ID until completed.
Kling 4.0 generates continuous video clips from 3 to 30 seconds in a single pass.
Omni Reference supports up to 10 images and 5 videos (up to 15 references in total) per request.
Yes, Kling 4.0 produces synchronized environmental audio and speech lip-sync, or can keep original audio from an input video.
720p and 1080p are supported with auto, 16:9, 9:16, and 1:1 aspect ratios.