Kling 4.0: Text, Image, Reference and Video Edit in One AI Video Model

1 Text to video: start from a sentence

Describe the scene and get a clip of 3 to 30 seconds in 16:9, 9:16, 1:1 or 21:9. It is the fastest way to try an idea, and with 9:16 the result is ready for Reels, Stories and TikTok without cropping.

720p, 9:16, three seconds, from the prompt "A red vintage convertible drives along a coastal road at golden hour, cinematic tracking shot." Subject, action, place, light and one camera move: that short recipe is enough for a usable first take.

2 Image to video: animate a frame you already have

Upload a start image, or pick one from your image history, and describe what happens next. Add an optional end image and Kling 4.0 moves from the first frame to the last: a product turning into its packaging, a character walking from one room into the next. This tab has no aspect ratio setting, so start from an image in the format you want to post.

3 Reference to video: your cast, your moves

Give Kling 4.0 up to 10 reference images and up to 5 reference videos (3 to 30 seconds each, 30 seconds combined) and combine them in one new scene. A character sheet, a product, a location photo and a clip with the motion you want can all go into the same generation. Pick an aspect ratio, or leave it on Auto to take the format from your references. Images and videos from your history work as references too.

4 Edit: change a clip, keep the rest

Upload a video of 3 to 30 seconds (MP4 or MOV, up to 100 MB, at least 700 pixels wide and high) or pick one from your history, and say what should change: the background, the weather, an outfit, the time of day. Add reference images if the change has to look like something specific. There is no duration to set: the result is as long as your source.

Call your references by number

Images are @Image1, @Image2 and so on, videos are @Video1, @Video2, in the order you added them. Use those names in the prompt and give each reference a job:

"The woman in @Image1 walks through the street in @Video1, wearing the coat from @Image2. Keep her face and hairstyle the same throughout."

In an edit, say what must stay

In the Edit tab your video is @Video1 and any reference images are @Image1, @Image2. Write one direct instruction, then name what must not change, so the model does not touch more than you asked for:

"Replace the jacket of the man in @Video1 with the jacket in @Image1. Keep his face, his movements, the camera and the background unchanged."

Animate the image, don't describe it again

In the Image tab the start image already shows what the scene looks like. Spend the prompt on what happens: how the subject moves, what changes around it, what the camera does. With an end frame, describe the transition between the two.

One camera move per shot, beats for long clips

Pick one move per shot (a dolly-in, a tracking shot, an orbit or a locked-off camera) and give characters a fixed label instead of "he" or "she". For a clip of 15 or 30 seconds, write the action as a few clear beats in order. Not in the mood to write it yourself? The Prompt Agent in the studio knows Kling 4.0 and its four modes and writes the prompt with you, references included.

Up to 30 seconds in one take

Choose any length from 3 to 30 seconds. That is enough for a full Reel in one generation, without stitching short clips together. An edit always keeps the length of its source.

The cost is on the button

Kling 4.0 is billed per second, at a rate set by the resolution. With reference videos, or in the Edit tab, the seconds of video you put in count as well, at a lower rate per second and at least twice the output length; an edit is billed on twice the length of its source. The resolution menu shows the rate and the Generate button shows the exact amount before you start.

Straight from and to your history

Start frames, reference images, reference videos and the video to edit can all come from your history, so a still from this morning can be a reference this afternoon. Every result lands in your video history, ready to download, cut in the Video Editor or post with Social Post.