Kling 4.0

Kling AI’s newest video model: 3 to 30 seconds with sound in every clip, frames, references and video editing.

Whisperer's verdict: 4.4 / 5

Kling 4.0 is Kling AI’s newest video model, made by Kuaishou. It makes a clip of 3 to 30 seconds from a prompt, a first frame or a set of references, in 720p or 1080p, and the sound comes out in the same run: dialogue, singing, sound effects and ambience. It can also edit a video you already have. This guide shows how each of its four modes works in the Lab, and how to write prompts that get what you want.

• What it doesVideo with sound from words, frames or references; edits your own clips
• Best forLong takes, short stories, ads, characters that talk
• LengthAny length from 3 to 30 seconds
• Resolution720p or 1080p
• ReferencesUp to 10 pictures and 5 videos
• Cost in the LabPer run, shown on the button before you click

Where to find it

1. Open the Video Generator — and pick Kling AI.
2. Choose Kling 4.0 — the version marked New, next to Kling 3.0.
3. Pick a mode — Words, Picture, References or Edit a video, then set the length (3 to 30 seconds) and the quality (720p or 1080p).

Sound is always on with Kling 4.0, so there is no sound switch: you steer it with words in the prompt. The price on the Generate button follows the length and the quality you choose, and with a video added it also counts the length of your videos.

The four modes side by side

Words: a scene from a description

Describe the subject, the action, the camera, the look and the sound. Choose the shape: 21:9, 16:9, 9:16 or 1:1. Kling 4.0 expands a short idea into a scene, and it follows a detailed shot brief line by line.

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.

Picture: start from your own frame

Add a first frame and Kling 4.0 starts the video from it. You can add up to 8 middle frames, in playing order, and an end frame: 10 frames in all. The video takes the shape of the first frame, and the prompt is optional here. Spend it on what happens after the frame, not on describing the picture again.

1. First frame — the exact opening picture.
2. Middle frames — only for moments that must be fixed: a character’s state, a scene change, a reveal.
3. End frame — where the shot has to land.
4. Prompt — the movement that connects the frames.

References: up to 10 pictures and 5 videos

Add pictures and videos, then name them in the prompt as @Image1, @Image2 and @Video1, numbered in the order you add them (tap a name in the panel to drop it into the prompt). A picture can set a character, a product, a place or a storyboard; a video can set the movement, the camera path and the pacing. Choose a shape, or Auto to let Kling 4.0 pick it from what you added.

@Image1 is an eight-panel storyboard grid. Animate it as one continuous shot that moves through every panel in order: a lion dance troupe leaps across tall pillars above a rooftop city at sunset, with drums and crowd cheers.

@Image1 is the character sheet for a clay astronaut dog and @Image2 is the storyboard. Follow the storyboard panel by panel as a stop-motion claymation short on a cheese moon, keeping the dog on model in every shot.

Use the camera movement from @Video1. The woman in @Image1 runs through the white office corridor in @Image2, with the camera tracking close behind her.

Edit a video: change one thing, keep the rest

Upload the clip you want to change: it becomes @Video1, and any extra reference videos follow from @Video2. Pictures you add are @Image1, @Image2 and so on. Say what to change and what must stay the same. The new video is as long as you choose, not as long as your clip.

Remove the crowd of passers-by from @Video1, leaving only the band and the fountain plaza. Keep the music and the camera unchanged.

Eight rules for better results

1. Give each reference one job — write the assignments before the action: “@Image1 is the character. @Image2 is the room. @Video1 sets the camera path only.”
2. Say what changes, then what stays fixed — a picture already shows the look; spend the words on the movement, and protect what matters: “keep the label unchanged”.
3. Give a sequence a beginning, a middle and an end — set the subject up, show the key action, and leave a short hold at the end. If it feels rushed, remove an event.
4. Keep the camera physically clear — name the opening framing, the path and the final view, and never ask for a locked camera and an orbit in the same shot.
5. Write dialogue as dialogue — quote the line, name the speaker, then the language and the delivery. Write on-screen words exactly as they should read.
6. Keep edits narrow — put the change first, then protect the timing, the camera path and everything that should not move.
7. Use frames for real changes — add a middle frame only when a moment must be fixed, and describe the move between frames.
8. Change one thing at a time — keep each result with its prompt and settings; if the face drifts, fix the character reference; if the pacing drifts, simplify the timeline.

Kling 4.0 or Kling 3.0?

Kling 4.0 goes up to 30 seconds (Kling 3.0 up to 15), always makes sound, and adds 21:9, references and video editing. Kling 3.0 stays in the studio with its own strengths, including 4K and Motion Control. Draft a short take first, look at what changed, then re-run the best one longer or at 1080p.

Questions