KLING AI · MODEL GUIDE
Every frame. Your direction.
Build a scene with realistic motion, synchronized sound and a story that has room to unfold. Explore 30-second generation, 10 keyframes and the creative control of Kling 4.0.
Kling 4.0 is the next generation of the Kling AI video model family. It turns text or images into video, combines visual and voice references, and lets creators plan individual moments with multiple keyframes. Its focus is a more coherent scene: believable movement, stable subjects and sound that follows the performance.
Five creative directions to explore. Watch each video example alongside its reference input.
01 / 05
Keep fast action and expressive performances readable, even when the camera moves.
Tracking and orbit shots, subtle facial expressions, lip-sync and two-channel stereo bring the visual performance and its sound into the same scene.
02 / 05
Guide the subject, movement and visual style with Omni Reference.
Combine up to 15 items across images, videos, voice references and subjects. The individual limits include 10 images, 5 videos totaling 30 seconds and 7 subjects; these share the overall 15-item budget.
03 / 05
Plan important visual moments instead of leaving every transition to chance.
Single-generation clips span 3–30 seconds. Up to 10 keyframe images guide character states, scene changes and story beats. Longer action can unfold in one continuous take.
04 / 05
Explore product films, short dramas, animation and campaigns for different audiences.
Covers a wider range of visual styles, multilingual dialogue and readable in-scene text. Use references to keep product identity and composition aligned while exploring a new style.
05 / 05
A separate Kling 4.0 variant designed for frequent creative iteration.
Flash is announced with 3–20 second output at 720p and 8-bit SDR. It supports text-to-video, first-frame image-to-video and Omni Reference. The full model’s 4K and 10-keyframe specifications should not be assumed for Flash.
| Capability | Kling 4.0 | Kling 4.0 Flash |
|---|---|---|
| Video duration | 3–30 seconds per generation | 3–20 seconds per generation |
| Output resolution | 720p / 1080p (4K coming soon) | 720p |
| Text-to-video | Supported | Supported |
| Image-to-video | First & last frames | First frame |
| Multi-keyframes | Up to 10 keyframe images | Not specified in the Flash table |
| Omni Reference | Supported; up to 15 combined items | Supported; limits not specified in the Flash table |
| Aspect ratios | 21:9 / 16:9 / 1:1 / 9:16 / Auto | Not specified in the Flash table |
| Audio output | Two-channel stereo; voice reference input | Not specified in the Flash table |
| Prompt input | Up to 8,000 tokens | Not specified in the Flash table |
| Video editing | Up to 5 clips, one primary; 3–30s each, ≤30s combined | Not specified in the Flash table |
Kling 4.0 is live on Seedance 2. Pick text-to-video, image-to-video or reference-led video and start creating with your prompt, images and references.
Start with a scene, one clear action, a camera move and the sound you want. Useful for original shots and early story exploration.
Create nowBring a product photograph, portrait or illustration. Describe what moves and what should remain unchanged.
Create nowPlan visual and motion references to keep your subject consistent across the scene.
Create nowDescribe the subject, action, setting, camera and audio. For a long take, define the beginning, middle and ending. These are original example prompts, not claims about generated results.
01
A matte-black perfume bottle on a wet stone plinth at dawn. Start in a close-up of water droplets, slowly orbit the bottle, then pull back to reveal the coastline. Keep the label unchanged. One continuous shot, soft side light, natural ocean ambience, no spoken dialogue.
02
Two friends meet outside a quiet café at sunset. The woman stops, smiles and says, “I thought you had left.” The man turns and replies, “Not without saying goodbye.” A gentle push-in, natural pauses, consistent faces, street ambience and clearly synchronized speech.
03
Begin with the explorer at the entrance of a glass greenhouse. In the middle, follow her as she discovers a glowing plant. End with a close-up of her surprised expression. Match the supplied keyframe images and outfit, keep the action continuous, and use a smooth tracking camera.
Start with the control your scene needs. Kling 4.0 stands out for realistic motion, synchronized sound and reference-led direction; Seedance 2.5 offers a different reference and editing workflow. Both are available in our studio, so try the same prompt on each and pick the one that fits.
Kling4.0, Kling4 and Kling AI 4.0 all refer to Kling 4.0, the latest generation of the Kling AI video model. It is now live on Seedance 2 for text-to-video, image-to-video and reference-to-video.
Kling 4.0 has been released and is already available on Seedance 2. Open the studio and select Kling 4.0 to start creating.
Yes. Kling 4.0 is live on this website. Members can select Kling 4.0 in the studio for text-to-video, image-to-video and reference-to-video.
720p and 1080p are already supported. 4K is coming soon.
Each generation can be 3–30 seconds long. Choose the duration in the studio before you generate.
Keyframes fix important moments in a video. In image-to-video on Seedance 2, upload a first frame and an optional last frame to set where the shot starts and ends.
In reference-to-video, you can combine up to 10 images and 5 videos, with no more than 15 items in total. Each reference video can be 2–30 seconds long, and reference video length also counts toward the billed duration.
Yes. Without a reference video, Kling 4.0 generates native audio automatically, with dialogue synced to the characters’ lips. Write the lines, tone and ambient sound into your prompt.
Kling 4.0 Flash is a lighter variant built for fast iteration, with clips up to 20 seconds at 720p. Seedance 2 offers the full Kling 4.0 model, with clips up to 30 seconds at up to 1080p.
Kling 4.0 is a members-only model on Seedance 2. Generations use credits; the cost depends on duration, resolution and whether you use reference videos, and the studio shows the exact amount before you generate. See the pricing page for membership plans.
Seedance 2 currently supports reference-to-video: upload existing videos and images as references to generate a new clip. Direct editing of an existing video is not available yet.
Kling 4.0 is live on Seedance 2. Bring your prompt, images and references, and start creating now.
Create now
