- Kling 4.0 is now in the video model picker: any length from 3 to 30 seconds, 720p or 1080p, with built-in stereo sound. With no inputs it's text-to-video, in 16:9, 9:16, 1:1 or 21:9.
- Keyframes: add images in time order — the first is the first frame, the last is the last frame, and the ones in between are keyframes, up to 10 in total. How many you can use depends on length: 4 at 3 seconds, one more per extra second, and 10 from 9 seconds up. Extra frames are greyed out — tap "Extend to N s" to use them. With a first frame, the aspect ratio follows it.
- Open ending: turn it on and the last image becomes a keyframe too, leaving the ending to the model.
- Prompt per frame: one click inserts an outline — "opens on @FirstFrame → then to @Keyframe → @LastFrame" — so you just fill in the action for each part.
- Multimodal Ref: use images, videos and audio together as references — up to 10 images, 5 videos (2–30 s each, 30 s combined) and 3 audio clips, 15 items in total. Mention them in the prompt with @.
- Pricing: 5 seconds costs 95 credits at 720p and 142 at 1080p. With a reference video it's billed under Kling's official rules, and you'll see the price before you generate.
- CLI and MCP too: update to the latest version and add keyframes in time order with
--keyframe-image.
changelogOct 23, 2026