Plan your Kling 4.0 video with prompts and model controls for 30-second videos, up to 10 keyframes, Omni Reference and stereo audio. Explore official samples alongside the workspace.
Kling4AI is not affiliated with Kling AI or Kuaishou.
Kling 4.0 controls · preview
Describe the shot. Prompts can run up to 8,000 tokens.
Video generation is currently unavailable. Changing these controls does not submit a job, and you will not be charged. You can also use Kling AI.
A 7-second reunion, directed in one paragraph. This official clip uses the sample prompt initially shown on the left. Changing the controls does not generate a new result. 3 s excerpt of a 7 s clip.
Official samples
Short excerpts of the clips Kling AI published, grouped the way its release note groups them. Nothing loads until you press play, and only one clip plays at a time.
An action scene in an ornate room, framed in 21:9. Kling says high-speed movement, continuous action, and tracking or orbit shots stay stable and seamless.
Handle large-scale scenes with ease · Reference image · with audio
A vlog-style introduction. Kling points to facial detail, lighting, texture and more accurate lip sync as the improvements.
Lifelike visuals, down to the details · Prompt · with audio
Kling says Kling 4.0 can expand and refine simple prompts “with a director’s eye”. This is the shortest prompt in the release note.
Smarter prompt understanding · Prompt · with audioOne paragraph sets the shot length, place, action, camera, look and sound. The original clip runs 7 seconds.
Smarter prompt understanding · Prompt · with audio
Omni Reference can borrow the pacing, shot composition and visual expression of an existing video while replacing the products or characters. This result comes from a reference video plus a reference image.
Effortless creative recreation · Reference video + reference image · no audio
Three reference images guide a sneaker commercial. Kling lists advertising next to filmmaking, short-form drama, animation and music videos. The original clip runs 30 seconds.
More possibilities for creative production · Product ad reference images · with audio
One of six style examples in the release note, next to vlog, realistic documentary, glitch, felt art and 3D cyberpunk.
A wider range of visual styles · Style example · with audio
From the Flash tier, which Kling describes as fast generation with excellent value for frequent creation. Flash makes 3–20 second videos at 720p.
Kling 4.0 Flash for faster, easier creation · Reference image · with audioCredit: Kling AI. Clips are short, re-encoded excerpts of the samples in the Kling 4.0 release note, verified October 23, 2026, shown here for reference.
What’s new in Kling 4.0
Kling AI frames the upgrade around realism, control and professional quality. These are the changes its release note lists.
Videos run from 3 to 30 seconds, long enough for long takes and continuous storytelling.
3–30 sMark key visual moments and specify character states, scene changes and story beats so the narrative unfolds as planned.
Multi-keyframesMix images, videos and subjects in a single generation. A subject can carry a specified voice.
Omni ReferenceHigh-speed movement, continuous action, tracking shots and orbit shots stay smooth and seamless.
Camera and motionAudio comes out in two-channel stereo, and more accurate lip sync keeps voices, dialogue and performances in step.
AudioOutput at 720p, 1080p or 4K, with a new 21:9 ultrawide format. 10-bit HDR for 1080p and 4K is coming soon.
OutputLegible on-screen text in Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, plus emoji.
MultilingualSimple prompts are expanded and refined “with a director’s eye”, and long, carefully crafted prompts are supported too.
PromptingOmni Reference
Omni Reference steers a generation with up to 15 items. Image references can be subject images, three-view references, image grids, infographics, keyframes or wireframes. Video references can carry a performance, an action, a camera move or the narrative pacing, and white-model and depth-map videos work too.
See Kling AI’s Omni Reference examples ↗Limits per Kling AI: 10 images, 5 videos (30 seconds combined), 7 subjects (up to 3 video-based) and voice references, 15 items in total.
Ways to start
The capabilities Kling AI lists for Kling 4.0, from a plain prompt to editing an existing clip.
Describe the shot. Prompts can run up to 8,000 tokens.
Start from an image and carry it into motion.
Set how the shot opens and how it ends.
Up to 10 keyframe images mark the beats in between.
Up to 15 images, videos, subjects and voices guide identity, motion and camera.
Up to 5 input clips, one of them the primary video, for targeted edits.
Official prompts
Three prompts from the release note, quoted exactly, each next to the clip it produced.
A vlogger-style video in which the woman cheerfully introduces herself to the audience. Maintain a natural, authentic look and realistic image quality.
An old man ranting about AI.
A continuous 7-second shot set in the UK. Two friends who have not seen each other for a long time meet at the corner of a quiet street at dusk, surrounded by a diverse crowd going about their daily lives. They smile, walk quickly toward each other, and share a warm embrace. Their facial expressions and body movements are natural. A handheld camera follows them with subtle camera shake. Soft, warm lighting with a nostalgic cinematic film look. Include natural footsteps, street ambience, background music, and English dialogue.
Anatomy of a long prompt
The longest prompt in the release note puts the shot length, place, action, camera, look and sound into one paragraph.
Specifications
Taken from the capability tables in Kling AI’s release note. “Not listed” means the note does not list that capability for Flash.
| Capability | Kling 4.0 | Kling 4.0 Flash |
|---|---|---|
| Text-to-video | Yes | Yes |
| Image-to-video | Yes | Yes, from a first frame |
| First & last frames | Yes | Not listed |
| Multi-keyframes | Up to 10 keyframe images | Not listed |
| Omni Reference | Yes | Yes |
| Video duration | 3–30 seconds | 3–20 seconds |
| Resolution | 720p / 1080p / 4K | 720p |
| Dynamic range | 10-bit HDR at 1080p and 4K (coming soon) | 8-bit SDR |
| Kling 4.0 spec sheet | Details |
|---|---|
| Aspect ratio | 21:9 / 16:9 / 1:1 / 9:16 / Auto |
| Audio output | Two-channel stereo |
| Languages | Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, Hindi and more |
| Dialects and accents | Chinese: Beijing, Taiwanese Mandarin, Northeastern, Sichuan dialect, Cantonese and more. English: American, British, Indian and more |
| Text generation | Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, Hindi and emoji |
| Text input length | Up to 8,000 tokens |
| Reference images | Up to 10 images |
| Reference videos | Up to 5 videos, up to 30 seconds combined |
| Reference audio | Voice reference only |
| Reference subjects | Up to 7 subjects (elements), including up to 3 video-based |
| Reference total | Up to 15 items across images, videos, voice references and subjects |
| Video editing | Up to 5 input clips including 1 primary video. Each clip 3–30 seconds, 30 seconds combined |
Source: Kling AI release note, “Kling 4.0 Core Capabilities and Specifications”, verified October 23, 2026. Specifications may change.
Alongside the model
Kling AI also rebuilt the page where videos are made. These are the four changes it describes.
Images, videos and audio references now live in a single input box instead of a toolbar.
Browse and compare creations in a grid, or review and edit them one by one in a list.
Preview, edit and add assets on a single timeline, and insert new material at any point.
A node canvas connects assets, process and results, and an Agent can use tools to carry out a request.
Availability
Kling AI’s release note confirms the model became available to everyone on its platform on October 15.
Kling AI’s MCP and CLI now support Kling 4.0, including Omni Reference and video generation with keyframes. These integrations are not connected to Kling4AI.
10-bit HDR output at 4K and 1080p, and video extension up to 2 minutes, are both listed as coming soon.
Good to know
Short answers drawn from Kling AI’s own release note.
Kling 4.0 is Kling AI’s video model. Its release note highlights 3–30 second generations, up to 10 keyframes, Omni Reference with up to 15 inputs, two-channel stereo audio and output up to 4K.
Kling 4.0 became available to everyone on Kling AI’s platform on October 15, 2026. Its release note, verified on October 23, also confirms MCP and CLI support for Omni Reference and video generation with keyframes. Generation is not available on Kling4AI.
Between 3 and 30 seconds in a single generation. Video extension, which Kling AI says can extend a video multiple times up to 2 minutes, is listed as coming soon.
720p, 1080p and 4K, in 21:9, 16:9, 1:1, 9:16 or Auto. 10-bit HDR output for 1080p and 4K is listed as coming soon.
Yes. Audio is output in two-channel stereo, with more accurate lip sync. Supported languages include Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Hindi, with accents and dialects such as Sichuan dialect, Cantonese, and American, British and Indian English.
Omni Reference accepts up to 15 items per generation: up to 10 images, 5 videos (30 seconds combined), 7 subjects (up to 3 of them video-based) and voice references. Multi-keyframes accepts up to 10 keyframe images.
Kling 4.0 Flash is the variant built for frequent creation, combining fast generation with excellent value. It supports text-to-video, image-to-video from a first frame and Omni Reference, with 3–20 second videos at 720p in 8-bit SDR.
Yes. Video editing accepts up to 5 input clips, including one primary video. Each clip must be 3–30 seconds and all clips together must not exceed 30 seconds. Kling AI says edits can target expressions, body movement, camera position and movement, style and background while the rest of the scene stays stable.
Kling 4.0 is available on Kling AI’s platform at kling.ai and through its MCP and CLI integrations. Kling4AI is an independent video-generation platform with a workspace preview; no generation provider is connected here. You can edit a prompt, explore model controls and view official sample clips, but you cannot submit generation jobs or pay for generation here. Kling4AI is not affiliated with Kling AI or Kuaishou.