Kling 4.0 doubles native video length to 30 seconds
Kuaishou's Kling 4.0 doubles single-pass video generation to 30 seconds and adds up to 10 keyframes, giving creators more control over timing and motion in longer AI-generated clips.
Kling has published a comparison of its new Kling 4.0 video model and the Kling 3.0 series. The main change is length: Kling 4.0 generates clips of 3 to 30 seconds in a single pass, up from 3 to 15 seconds in 3.0. The company says the model is built to keep the narrative consistent across the longer clips.
Kling 3.0 let creators set only a start frame and an end frame. Kling 4.0 accepts up to 10 keyframe images, so creators can fix the visual moments at points along the timeline. Reference inputs also grow. A single generation can now use up to 15 combined assets: up to 10 images, 7 subjects, 3 video-based subjects and 5 reference videos totalling 30 seconds, plus voice references. Kling 3.0 allowed at most seven images or subjects, or one video plus four images or subjects.
Other changes in the spec table:
- 10-bit HDR output at 1080p and 4K, up from 8-bit SDR
- Two-channel stereo audio, up from single-channel
- New languages added to Chinese, English, Japanese, Korean and Spanish, including Portuguese, German, French and Hindi
- Video editing built into the main Kling 4.0 workflow instead of running through separate 3.0 Omni workflows
Kling also says the model handles fast action, multiple moving subjects and large camera moves such as tracking and orbit shots more steadily. It gave no benchmarks or independent evaluations to back up these claims.
According to the post, Kling 4.0 was due to launch officially in October. A lighter variant, Kling 4.0 Flash, opened to a limited group of early-access users on September 28. The post does not mention pricing or API availability.