Storytelling with keyframes
Set the first frame, up to 8 middle beats and the last frame. Kling 4.0 invents the motion between your storyboard beats while keeping continuity.
Turn prompts, keyframe storyboards and up to 15 reference materials into realistic AI video with synchronized stereo audio — clips from 3 to 30 seconds at 720p or 1080p.
Describe any scene and Kling 4.0 generates it with synchronized audio. No uploads required.
Kuaishou’s flagship video model built for professional creation — realistic motion, synchronized stereo audio and two directed workflows: multi-keyframe storyboards and omni reference.
From a single sentence to a fully referenced scene, Kling 4.0 keeps motion, audio and identity under your direction.
Set the first frame, up to 8 middle beats and the last frame. Kling 4.0 invents the motion between your storyboard beats while keeping continuity.
Mix reference images, videos and voices in one request so characters, products and styles stay consistent across the shot.
Native two-channel sound with improved lip sync is generated in the same pass — or keep the original audio of your reference videos.
Longer single-shot storytelling with continuous action, narrative development and stable camera movement.
High-speed movement, tracking shots, orbits and multi-layer scenes stay smooth and stable.
16:9, 9:16, 1:1, 21:9 or auto — let the output follow your reference materials or pick the exact ratio you need.
Pick a workflow, attach your materials and generate.
Text to Video for pure prompts, Keyframes for storyboard control, or References to lock characters, products and voices.
Upload frames or reference media, then mention each item in the prompt with @ — every mention must match an attachment.
Set resolution, duration and aspect ratio. The studio estimates your cost live, including reference video input duration.
Rendering takes a few minutes. Preview the result inline and download the finished MP4.
Answers to common questions about modes, materials and pricing.
Kling 4.0 is Kuaishou’s fourth-generation AI video model. It delivers realistic motion, synchronized stereo audio and strong prompt understanding, with clips from 3 to 30 seconds at 720p or 1080p.
Keyframes mode animates a storyboard: upload a first frame, up to 8 middle keyframes and a last frame, then describe the journey in the prompt. References mode feeds the model up to 15 reference items — images, videos, voices and subjects — to keep characters, products and styles consistent.
Type @ in the prompt to mention attachments. Keyframes mode uses @first_frame, @keyframe_1… and @last_frame. References mode uses @image_1, @video_1… and the custom name of each voice. Every @ mention must match an attached asset.
Up to 8 middle keyframes, and the total number of frames (first + middle + last) is capped by duration: 4 frames at 3s, 6 frames at 5s, and 10 frames at 9 seconds or longer.
Up to 10 images, 5 videos (each 2–30 seconds, 30 seconds combined), 3 voices and 7 subjects — 15 items in total across all four types.
Without video input (text, images, keyframes or voices only) you pay the no-video rate × output duration: 40 gems/s at 720p and 80 gems/s at 1080p. With video input the rate drops to 20 gems/s at 720p and 40 gems/s at 1080p, and input duration is billed too: rate × (output × 2) when total input ≤ output, otherwise rate × (input + output). For example, a 10-second 720p output with a 5-second input video costs 20 × 20 = 400 gems.
Yes. Kling 4.0 generates synchronized native audio by default. When you attach reference videos you can choose to keep their original sound instead.
Start with a prompt, a storyboard or a set of references and generate cinematic video with Kling 4.0.
Start creating with Kling 4.0