Start Frame

Click or drag to upload an asset

Supports JPG, PNG, and WEBP. Max 10MB each.

Middle Keyframes

Up to 6 · ≤ 10MB

Add Middle Keyframes

End Frame

Click or drag to upload an asset

Supports JPG, PNG, and WEBP. Max 10MB each.

Prompt

Required

0/8000

Duration

3s30s

Kling 4.0 AI Video Generator: References & 30s

Generate up to 30 seconds of video with native audio using Kling 4.0, the latest AI video model from Kling AI. Turn a prompt, up to 15 reference images, videos and voices, or a set of keyframes into multi-shot clips at 720p or 1080p.

Kling 4.0 video of a rainy neon night market where a woman in a yellow raincoat orders from a street cook

A 10-second cinematic short in three shots with hard cuts. Shot 1: a wide crane shot descends over a rain-soaked night market in an old Asian city, neon signs reflecting in puddles, steam rising from food stalls. Shot 2: close-up of an elderly street cook tossing noodles in a flaming wok. Shot 3: a young woman in a yellow raincoat lowers her umbrella, smiles at the cook and says softly, "Same as always, please." 35mm film look. Audio: rain on tarps, wok sizzle, crowd chatter and her spoken line.

How to use Kling 4.0

Move through prompt, reference images, and model selection in the same order you generate.

Choose a tab

Open the Kling 4.0 AI video generator and pick a tab. Text to Video needs only a prompt. Image to Video takes a first frame, optional middle keyframes and a last frame. Multimodal takes reference images, videos and voice clips.

Write the prompt

Describe the shots, camera moves and dialogue in up to 8,000 characters. Call each upload by the name on its thumbnail, such as @image_1, @video_1, @voice_1 or @first_frame, @keyframe_1 and @last_frame, so the model knows what every file is for.

Set length and generate

Pick 3 to 30 seconds, 720p or 1080p and a 16:9, 9:16, 1:1 or 21:9 frame. Generate, watch the clip with its sound, then adjust the prompt or swap a frame and run it again.

Why choose Kling 4.0

Explore the features that help you create clearer, more consistent images.

Kling 4.0 multi-shot night market video with neon reflections, a flaming wok and spoken dialogue

30-second multi-shot clips with sound

Kling 4.0 generates up to 30 seconds in one pass, twice the length of Kling 3.0. That is enough for a full scene with several shots and hard cuts. Write the shot list in your prompt and get stereo dialogue, ambience and effects in sync with the picture.

Generate a 30s video

Kling 4.0 perfume ad made from reference images, with the same model and bottle in every shot

Reference to video with consistent characters

Add up to 10 reference images and 5 reference videos, then point to them in the prompt as @image_1 or @video_1. Faces, products and locations stay the same from shot to shot, which makes it practical for ads, series and branded content.

Use your references

Kling 4.0 keyframe video of a woman going from a tram stop through a flower market to a rooftop party

Keyframe to video: first, middle and last frames

Upload a start frame, an end frame and up to 8 frames in between. The model moves through them in order, so you decide where the story starts, which places it passes and where it lands. Longer clips allow more frames, up to 10 in total.

Try keyframe to video

Lip-synced dialogue scene in a rainy bookshop where an old bookseller hands a boy a notebook

Voice references and lip-synced dialogue

Upload up to 3 voice clips and give a character that voice with @voice_1. Spoken lines are lip-synced to the faces on screen. When you add a reference video, you can keep its original sound instead of generating new audio.

Add a voice

FAQ

Common decisions to check before generating.

Kling 4.0 is the latest AI video model from Kling AI, the video platform of Kuaishou, released in October 2026. It creates 3 to 30 second videos with native audio from text, keyframes or reference images and videos. On Pic Creator it renders at 720p or 1080p.

Kling 3.0 tops out at 15 seconds per clip. Kling 4.0 doubles that to 30 seconds, lets you place up to 10 keyframes in one video, accepts up to 15 reference files per task and handles camera moves such as orbits and tracking shots more steadily.

Open the Multimodal tab and upload up to 10 images, 5 videos and 3 voice clips. Each file shows its name on the thumbnail, such as @image_1, @video_1 or @voice_1. Use those names in the prompt, for example "the woman in @image_1 picks up the bottle". Every @ must match an upload, or the task cannot start.

Open the Image to Video tab, upload the first frame, add middle keyframes in story order and finish with the last frame. A 3-second clip takes up to 4 frames and clips of 9 seconds or more take up to 10. Describe the motion between them and call each frame @first_frame, @keyframe_1 or @last_frame.

Pic Creator renders 720p or 1080p MP4 clips from 3 to 30 seconds, in 16:9, 9:16, 1:1 or 21:9. With reference files it can also follow the shape of your uploads. Kling has announced 4K output, but it is not available yet.

Creator reviews

Feedback closer to real usage than marketing copy.

“We storyboard each ad as four keyframes: packshot, texture, model and end card. Kling 4.0 fills in the motion. Variants that used to take a day of editing now take an afternoon.”

Lauren Brooks

Ad creative lead

“A 30-second take with three cuts and the dialogue already in sync is enough to pitch a scene. I still regenerate a few times to get the performance right, but the shots hold together.”

Marcus Webb

Independent filmmaker

“Our packshot goes in as @image_1 and the model photo as @image_2, and the product looks the same in every frame. That consistency is why we moved our product videos to this model.”

Priya Nair

E-commerce content manager