Describe a scene or upload your images to create videos up to 30 seconds long. Showcase a product, bring a character into motion, or turn an ad concept into a video—with control over the action and camera movement.
Loading models…
Kling 4.0 is Kuaishou’s multimodal AI video generation model, combining video creation and targeted editing in one workflow. It emphasizes realistic character performance, stable motion, and cinematic camera work, with synchronized audio and lip sync.
Multiple keyframes guide a scene’s beginning, progression, and ending, while references shape characters, products, settings, and movement. Creators can also edit expressions, actions, backgrounds, or camera work in existing footage—for short films, product ads, animation, and character-led stories.
Develop a complete scene, adapt existing visual assets, or refine footage around a specific creative brief. Explore how narrative controls, audiovisual generation, and targeted editing support different stages of production.
Longer video generation and multiple keyframes let you plan how a sequence unfolds beyond its opening image. Set the moments that matter—a character’s arrival, a product reveal, or a final reaction—and describe the action connecting them. This gives a short story or advertisement room to develop, with greater control over pacing and how it ends.
Build Your Next Scene
Improved motion handling and camera control help depict moving subjects through tracking shots, orbits, and changes in framing. High-resolution HDR output helps preserve texture and tonal detail in scenes with bright highlights and deep shadows. Together, these capabilities support sports, performance, and cinematic footage where both the action and its surroundings matter.
Direct Your Camera Moves
Use images and video clips as references to guide character appearance, product details, settings, and movement. Bring a product into a different environment, build a scene around an established character, or draw on a reference clip’s pacing and composition. Develop new campaign variations using existing assets to guide your brand’s visual identity.
Create from Your References
Native audio generation creates speech, ambience, and sound effects alongside the visuals, with stereo sound and lip sync connecting them to the picture. Describe a character’s lines, the surrounding sounds, and the atmosphere you want; voice references can also guide how the speaker sounds. Easily add character voiceovers for conversations, background noise for settings, or sound effects for action scenes.
Generate Video with Sound
Targeted editing lets you modify expressions, actions, backgrounds, camera work, or visual style in existing footage. Specify the adjustment and what you want to preserve, whether you need a different reaction, a new setting, or another visual treatment. Use specific revision notes to turn feedback into alternative versions of the scene.
Refine Your Existing Video
Generate dialogue in multiple languages, with regional options including Cantonese, Sichuanese, and American, British, or Indian English accents. Write the lines and specify their delivery to shape a regional conversation, a character’s voice, or an ad for another market. Support for multilingual on-screen text, emojis, and logos also helps you adapt visual elements for your audience.
Create in Your Language
Follow these three simple steps to create a video from your idea or reference materials, ready to download and share.
Describe the subject, action, camera movement, and any sound you need. Upload optional references and specify their roles.
Choose your duration, resolution, and aspect ratio, adding keyframes where timing matters. Check the credit cost, then generate.
Download your video for a short film, a larger film project, or social content, then share it with collaborators or publish it on your chosen platform.
HeyTop AI does not store your uploaded content or use it to train AI models. This applies to both personal and project materials.
You don’t need professional filming or editing skills to get started. The controls are approachable for beginners, with options for more detailed creative control.
Download high-resolution footage without visible platform watermarks. Use it in your own projects without an additional watermark-removal step.
Give your videos an instant high-budget look. Kling 4.0 delivers breathtaking visual fidelity with lifelike textures and complex lighting.
Give your ideas and existing materials a new role in video. The examples below show where the results can fit and how they can contribute to your project.
Club introductions, tournament announcements, and sports community posts can build anticipation around the energy of a sport. Promotional sequences offer organizers a way to introduce the event’s character and begin promotion before match footage becomes available.
Designers and fashion teams can complement seasonal lookbooks, portfolio presentations, and collection previews with moving imagery. Buyers and audiences gain another perspective on the styling direction, making it easier to discuss individual looks and the collection as a whole.
An existing concert, exhibition, or festival poster can become the visual foundation of a wider release campaign. Extending that artwork across social posts, event pages, and venue screens makes better use of completed design work while keeping the event recognizable.
Gallery previews, art explainers, and classroom discussions can use imagined scenes to draw attention to a painting’s composition and atmosphere. These interpretations provide a starting point for closer observation, encouraging viewers to discuss details they might otherwise overlook.
Tourism trainers can illustrate everyday encounters at visitor centers, hotel front desks, and guided tours, including directions and requests needing clarification. Learners get concrete situations to rehearse, while instructors can compare wording, explain cultural context, and discuss appropriate responses.
| Specification | Kling 4.0 | Kling 3.0 Series |
|---|---|---|
| Text to Video & Image to Video | ✅ | ✅ |
| Start & End Frames | ✅ | ✅ |
| Multiple Keyframes | Up to 10 keyframe images · NEW | Not supported |
| Video Duration | 3–30 seconds · NEW | 3–15 seconds |
| Resolution | 720p / 1080p / 4K | 720p / 1080p / 4K |
| Audio Output | Two-channel stereo · NEW | Single-channel audio |
| Reference Images & Subjects | Up to 10 images and 7 subjects, within the combined reference limit | Up to 7 combined without video; up to 4 with video |
| Reference Videos | Up to 5 videos, totaling up to 30 seconds | 1 video, up to 15 seconds |
| Combined References | Up to 15 assets across supported reference types | Up to 7 images/subjects, or 1 video + 4 images/subjects |
| Video Editing | Targeted editing within the multimodal workflow | Available through 3.0 Omni workflows |
✅ Supported · NEW indicates a new capability or expanded specification marked in the official Kling 4.0 release.
Reference limits apply together: individual input limits cannot be added beyond the combined maximum. Kling 3.0 capabilities vary by version, and its editing entry refers to 3.0 Omni. Available settings depend on the platform and creation mode.
You can start with text alone when you want the model to interpret an idea. Add references when a character, product, setting, or visual style needs a more specific starting point.
No. Use keyframes only where a particular visual moment matters, such as an entrance, transition, or ending. A simpler sequence may need just a few, or none.
Usage is charged in credits under HeyTop AI’s subscription plans. Visit the pricing page for plan allowances and check the credit cost shown before submitting each generation.
Describe the subject, setting, main action, and camera movement, adding dialogue or sound requirements where needed. When several events are involved, state their order and avoid conflicting instructions.
Use clear references of the same subject and identify the details that should stay consistent, such as clothing, shape, or color. Conflicting references can make the result less predictable, and exact reproduction is not guaranteed.
Targeted editing does not guarantee that every other detail will remain unchanged. Specify what to modify and what to preserve, then check the downloaded result for unintended changes.
