kwaivgi/kling-v4.0

Kling 4.0: generate 3 to 30 second videos at up to 1080p from text, keyframes, reference images, or reference videos, and edit existing videos

38 runs
Playground API Examples README

Kling 4.0

Kling 4.0 is Kuaishou’s latest video model. It generates videos from 3 to 30 seconds long at 720p or 1080p, and it takes almost any mix of inputs: a text prompt, keyframes, reference images, reference videos, or a video you want to edit.

Earlier Kling versions split these jobs across separate models (text-to-video, image-to-video, Omni). Kling 4.0 handles all of them in one model, and picks the right mode from the inputs you give it.

What you can do with it

Text to video. Describe a scene and get a video back. Pick the aspect ratio (16:9, 9:16, 1:1, or 21:9), the resolution, and a duration anywhere from 3 to 30 seconds.

Keyframes. Give it a start_image, an end_image, and any number of middle_images in between, and Kling 4.0 animates the path through them. The output takes its aspect ratio from your images.

Reference images. Pass up to 10 images of characters, products, places, or styles with reference_images, then point at them in your prompt as @image_1, @image_2, and so on. This is the way to keep a character or product looking the same across many clips.

Reference videos. Use reference_videos to borrow camera movement, motion, or style from existing footage. Refer to them in the prompt as @video_1, @video_2, and so on.

Video editing. Set base_video and describe the change, like “change the background of @video_1 to a tropical beach”. The output keeps the length and aspect ratio of the original video.

How references are numbered

Images are numbered in this order: start_image, then middle_images, then end_image, then reference_images. So if you set a start image and two reference images, the start image is @image_1 and the references are @image_2 and @image_3.

Videos are numbered with base_video first, then reference_videos.

Limits

  • Up to 10 images in total across all image inputs.
  • Up to 5 videos in total across base_video and reference_videos.
  • Each input video must be 3 to 30 seconds long, and all input videos together can’t add up to more than 30 seconds.
  • Input videos must be .mp4 or .mov, 700 to 4553 pixels per side, at most 300 MB, with an aspect ratio between 2:5 and 5:2.
  • Images can be up to 30 MB, with an aspect ratio between 2:5 and 5:2.
  • Prompts can be up to 26,000 English characters or 11,000 non-English characters.

Pricing

You pay per second of video:

  • Text, keyframe, or reference image inputs: $0.157 per second at 720p, $0.235 per second at 1080p.
  • With any video input (base_video or reference_videos): $0.094 per second at 720p, $0.141 per second at 1080p. Billed seconds are the output length plus whichever is longer, the input videos or the output. For example, editing a 5 second video into a 5 second output bills 10 seconds.

Tips for writing prompts

  • Describe the subject, the action, the setting, and the camera, in that order. “A red fox trots across a snowy meadow at sunrise, cinematic tracking shot” works better than “fox in snow”.
  • For longer videos, describe what happens over time: “she climbs the stairs, then steps out onto the balcony as the storm rolls in”.
  • When you use references, say what each one is for: “the woman from @image_1 walks into the room from @image_2”.

Privacy policy

https://app.klingai.com/global/dev/document-api/protocols/privacyPolicy

API terms

https://app.klingai.com/global/dev/document-api/protocols/paidServiceProtocol

Service level agreement

https://app.klingai.com/global/dev/document-api/protocols/paidLevelProtocol

Model created
Model updated