Loading model details…
A next-generation video model for professional audiovisual creation with native 30-second generation, 4K 10-bit HDR, stereo audio, multi-keyframes, and multimodal references.
Kling
Kling 4.0 is Kling AI's next-generation model for professional video creation, with major improvements in realism, control, audiovisual completeness, and long-form storytelling. It supports text-to-video, image-to-video, first and last frames, multi-keyframes, Omni Reference, and video editing, generating native clips from 3 to 30 seconds at 720p, 1080p, or 4K.
The release supports 10-bit HDR at 1080p and 4K, a 21:9 ultrawide format, and two-channel stereo while improving complex motion, camera work, lip sync, performance, and in-frame text. A request can use up to 10 keyframes and up to 15 combined Omni Reference assets, including as many as 10 images and 5 videos totaling no more than 30 seconds, plus voice and element references. The public material does not yet provide the final API model code or complete request fields, so integration parameters must follow the forthcoming API documentation and LinkModel's live model page.
| Feature | Benefit |
|---|---|
| 4K 10-Bit HDR | Preserves highlight, shadow, and color detail in backlit, night, and neon scenes for flexible finishing. |
| Native Generation up to 30 Seconds | Completes longer stories, advertisements, or explainers with stronger character, voice, and style continuity. |
| Up to 10 Keyframes | Anchors openings, turning points, and endings for more predictable cinematography and narrative control. |
| Up to 15 Omni References | Combines people, products, environments, motion, camera, style, and voice for multidimensional consistency. |
| 8,000-Token Text Input | Fits storyboards, timelines, dialogue, sound, and detailed directorial instructions. |
| Multilingual Stereo Audio | Generates dialogue, ambience, and effects across multiple languages, accents, and dialects. |
It supports text-to-video, image-to-video, first and last frames, multi-keyframes, Omni Reference, and editing of existing video. The disclosed material also describes repeated video extension for continuous content reaching approximately two minutes.
Native clips range from 3 to 30 seconds at 720p, 1080p, or 4K. The 1080p and 4K modes support 10-bit HDR, with 21:9, 16:9, 1:1, 9:16, and automatic aspect ratios.
One generation can use up to ten keyframes to anchor the opening, plot turns, action beats, and ending. Creators can vary keyframe count and placement and combine them with time-based descriptions for pacing and transitions.
One task can combine up to 15 image, video, voice, and element references, including up to 10 images and 5 videos with a combined duration of 30 seconds. The material also lists up to seven elements, including up to three video elements; current platform limits remain authoritative.
It generates two-channel stereo dialogue, ambience, and effects with improved lip sync. Disclosed languages include Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and Hindi, alongside Cantonese, Sichuan dialect, and multiple Chinese and English accents.
The model supports multilingual text, emoji, and logos with stronger readability during moving shots. Prompts should provide exact copy, language, placement, perspective, type character, and required screen time, followed by spelling and brand review.
Kling 4.0 is preparing for pre-listing on LinkModel and is not yet officially available for API calls. The pre-listing date, public model ID, request schema, and enabled capabilities should be confirmed through LinkModel's latest announcement and model page; promotional material should not be used to infer API fields.