Kling 4.0 is the fourth-generation Kling AI video model from Kuaishou, built for native 30-second clips, multi-keyframe storytelling, Omni references and consistent subjects, with stereo audio in 720p or 1080p.
Review, download and reuse videos generated with Kling 4.0.
Kling 4.0 · Model overview
Kling 4.0 is the fourth-generation video model from Kling AI, Kuaishou's AI video platform. It focuses on visual realism, creative control and narrative completeness: a single generation can run for 30 seconds, follow up to 10 keyframes and draw on up to 15 multimodal references, with stereo audio and output in 720p or 1080p.
Generate up to 30 seconds in a single pass, twice the 15-second limit of Kling 3.0, with room for an opening, the main action and a complete ending.
Place up to 10 keyframes along the timeline to lock composition, character positions and story beats, while Kling 4.0 generates the motion between them. A clip takes one more frame than its length in seconds, so a 3-second clip takes up to 4 and clips of 9 seconds or longer take 10.
Combine up to 10 images, 5 videos, 7 subjects and 3 voices, with no more than 15 references per generation, for more consistent characters, products and styles. Reference-based video editing is also enhanced.
Fast action, continuous movement and complex camera moves render more steadily and smoothly, which makes large-scale scenes, chases and sweeping shots easier to handle.
Render each generation in 720p for drafts and social feeds, or in 1080p for final delivery and footage that will be edited further. 4K and 10-bit HDR output are not offered here.
Sound is generated together with the picture, with high-quality stereo audio, more accurate lip sync and speech in multiple languages, accents and dialects.
Define up to 7 subjects from 2–4 images or one 3–8 second video, refer to them with @name in the prompt, and bind a voice recording to a character subject so that it speaks in that voice.
Key Kling 4.0 specifications at a glance.
| Specification | Kling 4.0 |
|---|---|
| Developer | Kling AI (Kuaishou Technology) |
| Versions | Kling 4.0 (full model, used here); a lighter Kling 4.0 Flash also exists |
| Single generation | 3 to 30 seconds, generated natively |
| Subjects | Up to 7, from 2–4 images or one 3–8 second video, with an optional voice |
| Keyframes | Up to 10 (one more than the clip length in seconds) |
| Omni references | Up to 15: 10 images, 5 videos, 7 subjects and 3 voices |
| Prompt length | Up to 8,000 characters |
| Resolution | 720p or 1080p |
| Aspect ratios | Smart, 16:9, 9:16, 1:1 and 21:9 ultra-wide |
| Audio | Dual-channel stereo with more accurate lip sync |
| Languages | Chinese, English, Japanese, Korean, Spanish and more, with accents and dialects |
Kling 4.0 comes in two versions. The workspace on this page runs the full model, built for realism and control; Flash is a lighter version tuned for speed and cost.
Full model for high-fidelity scenes
Lightweight version, not offered on this page
Use native 30-second generation and keyframes to stage a complete scene with an opening, rising action and a resolved ending.
Keep products and characters consistent with Omni references and named subjects, and deliver 1080p footage for launch pages, paid social campaigns and e-commerce listings.
Create explainers, character dialogue and localized versions with synchronized lip movement across multiple languages, accents and dialects.
Improved stability in fast motion, continuous action and complex camera moves suits chases, sports highlights and sweeping establishing shots.
Lock key shots with up to 10 keyframes to test composition, pacing and camera paths before committing to a full production.
Frame scenes in 21:9 ultra-wide and render in 1080p for cinematic openers, trailers and widescreen showcases.
Creation workflow
Kling 4.0 rewards planning. Define the story beats first, anchor them with keyframes and references, and then choose the output format that suits where the video will be published.
Step 01
Divide the clip into beats: for example, seconds 0–10 establish the subject and setting, seconds 10–20 carry the main action, and seconds 20–30 deliver the climax or final frame. A clear structure helps the model pace motion and camera work.
Step 02
In Image mode, upload the keyframes in playback order: the first is the first frame, the last is the last frame and the rest are middle frames. In Media mode, add image, video or subject references for characters, products and style. Kling 4.0 accepts up to 10 keyframes and up to 15 references.
Step 03
Describe the subjects, actions, camera movement, lighting, dialogue and sound. Prompts can run up to 8,000 characters, so complex scenes can be described shot by shot.
Step 04
Select an aspect ratio, including 21:9 widescreen, or leave it on Smart, then choose 720p for drafts or 1080p for the final version. Review the result and adjust keyframes or the prompt before the next take.
Kling 4.0 offers more control than earlier Kling models. These practices help turn that control into consistent, usable footage.
Place keyframes at the start and end of the clip to fix where the scene begins and lands, then add intermediate keyframes only where the story needs a specific moment.
Use one clean, well-lit reference per character or product so Omni Reference keeps each identity consistent instead of blending several looks.
Name the camera move for each part of the clip, such as a slow push-in, an orbit or a tracking shot, rather than giving one instruction for the whole 30 seconds.
Put spoken lines in quotation marks and state the speaker and the language or accent, so lip sync and voice match the right character.
List ambience, sound effects and music with their timing so the stereo audio supports the action instead of competing with it.
Give each subject a short, distinct name and refer to it with @name. Bind a voice to a character subject when its lines must sound the same from take to take.
The table compares the published capabilities of Kling 4.0, Kling 3.0 and Seedance 2.5 and shows where each model fits best.
| Capability | Kling 4.0 | Kling 3.0 | Seedance 2.5 |
|---|---|---|---|
| Single generation | 3 to 30 seconds, generated natively | 3–15 seconds | Up to 30 seconds |
| Story control | Up to 10 keyframes for timeline-level control | Multi-shot storytelling from a single prompt | First-frame control and 3D spatial pre-visualization for camera blocking |
| References | Up to 15: 10 images, 5 videos, 7 subjects and 3 voices | Kling 3.0 Omni: up to 7 images and 1 reference video | Up to 50: 30 images, 10 videos and 10 audio files |
| Resolution | 720p and 1080p | 720p and 1080p, plus a native 4K mode | 720p up to native 4K |
| Audio | Stereo audio, more accurate lip sync, multiple languages, accents and dialects | Native audio with multilingual speech and accents | Native audio, with audio files accepted as references |
| Best fit | Long, directed scenes that need keyframe precision, rich references and consistent subjects | Cinematic clips of up to 15 seconds for social media, ads and short scenes | Reference-heavy 30-second productions and 4K brand content |
Kling AI has released a major model every few months since June 2024. Kling 4.0 is the largest step in clip length, creative control and output quality.
| Version | Release | Highlights |
|---|---|---|
| Kling 1.0 | June 2024 | Public launch of Kling text-to-video and image-to-video generation |
| Kling 1.5 | September 2024 | 1080p high-definition output and improved motion quality |
| Kling 2.0 | April 2025 | Major gains in motion, prompt understanding and visual quality, plus multimodal video editing |
| Kling 2.5 Turbo | September 2025 | Faster, more cost-efficient generation with stronger prompt adherence |
| Kling 2.6 | December 2025 | First Kling model with native audio, including speech, sound effects and ambience |
| Kling 3.0 | February 2026 | 3–15 second clips, multi-shot storytelling and multilingual native audio |
| Kling 4.0 | 2026 | Native 30-second generation, up to 10 keyframes, 15 references and named subjects with voices |
Each example is written for a full 30-second clip, with clear beats, camera direction and sound so the model can pace the scene from beginning to end.
0–10s: A matte black smartwatch rotates slowly on a wet stone pedestal, soft rim light, macro lens. 10–20s: The camera orbits as the screen lights up and raindrops slide off the glass. 20–30s: Pull back to a wide shot of a rain-soaked rooftop at dusk, the watch face glowing. Deep ambient synth and soft rain in stereo.
A warm café at night. A young woman in a green coat looks up from her notebook and says in English, "You came back." The man across from her smiles and replies in Spanish, "Nunca me fui del todo." Slow push-in, shallow depth of field, quiet café chatter and soft piano in the background.
Keyframes 1, 5 and 10 as anchors: a courier on a red bicycle enters a narrow alley at sunrise, weaves through a crowded market at midday and stops at a harbor gate at sunset. Keep the outfit and bicycle consistent. Tracking camera at handlebar height, 21:9 widescreen, rising drum rhythm.
0–10s: A rally car drifts through a muddy forest corner, mud spraying toward the lens, low tracking shot. 10–20s: A drone shot rises and follows the car over a crest as sunlight flares through the trees. 20–30s: The car crosses the finish line in slow motion while the crowd cheers. Roaring engine and crowd noise in stereo.
A presenter in a bright studio holds a pair of wireless earbuds and says in English, "Twelve hours of battery, one small case." She repeats the line in Japanese, then smiles at the camera. Medium shot, soft key light, clean studio ambience.
21:9 widescreen. Dawn over a misty mountain valley. 0–10s: A slow aerial glide above the clouds. 10–20s: The camera descends toward a lone rider on a ridge. 20–30s: The rider stops as the sun breaks over the peaks. Rising orchestral music and gentle wind.
Answers about Kling 4.0 features, references, resolution, prompts and how it compares with other video models.
Kling 4.0 is a video generation model from Kling AI, Kuaishou's AI video platform. It generates native 30-second clips from text, images and multimodal references, supports up to 10 keyframes and 15 references, and adds stereo audio and more accurate lip sync. Generations are available in 720p and 1080p.
Kling 4.0 is billed in credits based on clip length and resolution, and reference videos add to the billed length. No download is needed, and the pricing page lists current plans and credit packs.
Kling 4.0 is the full model, with 3–30 second clips, first-frame, last-frame and multi-keyframe control, and up to 15 multimodal references. Kling 4.0 Flash is a lighter version tuned for faster, more cost-efficient generation, limited to 720p and clips of up to 20 seconds without keyframe control. The workspace on this page runs the full Kling 4.0.
A single Kling 4.0 generation can be any length from 3 to 30 seconds, in one-second steps. That is twice the 15-second limit of Kling 3.0, with room for an opening, the main action and a complete ending.
Generations here are rendered in 720p or 1080p; 4K and 10-bit HDR output are not offered. 720p suits drafts and social feeds, while 1080p suits final versions and footage that will be edited or color graded.
Omni Reference accepts up to 10 images, 5 videos, 7 subjects and 3 voices in one generation, with a combined maximum of 15 references. Multi-keyframe control accepts up to 10 keyframes, one more than the clip length in seconds: a 3-second clip takes up to 4, and clips of 9 seconds or longer take 10.
Omni Reference is the multimodal reference system in Kling 4.0. In one generation it accepts up to 10 images, 5 videos, 7 subjects and 3 voices, with no more than 15 references in total, to guide characters, products, style, motion and speech. Compared with earlier versions, it offers richer reference options and more consistent results.
Yes. Kling AI lists enhanced video editing as part of the Omni Reference upgrade in Kling 4.0. Video references, up to 5 per generation, can guide how an existing scene is reworked, while the prompt describes the change.
Kling 4.0 offers 16:9 widescreen, 9:16 vertical, 1:1 square and a 21:9 ultra-wide frame for cinematic compositions. Smart lets the model choose the frame from your inputs.
Kling 4.0 improves stability in high-speed motion, continuous action and complex camera movement. Large-scale scenes, chases and sweeping camera moves render more steadily and smoothly than in earlier versions.
Kling 4.0 generates high-quality stereo audio together with the video and improves lip-sync accuracy. Speech is supported in multiple languages, including Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, along with a range of accents and dialects.
Kling 4.0 upgrades realism, control and story length:
Both models generate clips of up to 30 seconds. Kling 4.0 stands out for keyframe-level timing control and named subjects that can carry their own voices. Seedance 2.5 accepts a larger stack of up to 50 references, including audio files, and outputs up to 4K, which suits reference-heavy brand and product work.
Kling 4.0 accepts prompts of up to 8,000 characters. The extra room allows a 30-second clip to be described shot by shot, including camera moves, dialogue and sound.
Structure the prompt around the timeline rather than a single moment:
Use keyframes and references for anything that must look exact, and use the text to direct motion and pacing.
Kling 4.0 is developed by Kling AI, the AI video generation platform of Kuaishou Technology. The Kling model family launched in June 2024 and has shipped major upgrades every few months, from Kling 1.0 to Kling 4.0.