Kling 4.0 Arrives in October: Features and Flash Early Access

Kling 4.0 and Flash: what is available?

The full Kling 4.0 release is planned for October; the announcement does not give an exact day. Flash is the early-access model for frequent creation. Its official specification lists text-to-video, image-to-video using a first frame, and Omni Reference, with 3–20-second output at 720p and 8-bit SDR.

The full model’s specification lists 3–30 seconds and 720p, 1080p or 4K. Those specifications should not be applied to Flash. Check the model and options shown in your Kling account before starting a generation.

Audio and visuals: motion, stereo sound and HDR

Kling describes steadier dynamic motion and complex camera movement, more realistic lighting and textures, high-quality two-channel stereo sound, and more accurate lip sync. These are announced improvements, rather than results from our own testing.

The release notes explicitly mark 10-bit HDR output at 1080p and 4K as coming soon. This distinction matters: an announced output format does not mean it is already selectable in Flash early access.

Omni Reference: up to 15 assets in one generation

Omni Reference combines images, videos, voice references and subject elements, with a total limit of 15 reference items. The detailed specification also sets individual limits: up to 10 images, 5 videos with a combined duration of 30 seconds, and 7 subject elements, including at most 3 video-based elements. Audio reference is for voices.

These limits sit within the total allowance; they are not separate quotas to add together. Kling also describes more consistent reference use and targeted edits to expressions, movement, camera position, style and backgrounds.

Storytelling: 30 seconds and 10 keyframes

Kling 4.0 is announced with native generation of up to 30 seconds in a single request and input of up to 10 keyframe images. Keyframes let creators specify visual moments, character states and scene changes along a narrative.

The two numbers describe different controls: 30 seconds is the full model’s single-generation duration ceiling, while 10 is the maximum number of keyframe inputs. Flash’s published duration ceiling is 20 seconds.

Languages, accents and video extension

The announcement expands support for languages, accents and dialects. The release notes include Chinese, English, Japanese, Korean, Spanish, Portuguese, German and French, with examples such as Cantonese, Sichuanese and American, British and Indian English accents.

Video extension is also announced. The detailed notes describe repeated extensions up to two minutes and mark the feature as coming soon. That is an extension workflow, not a two-minute native generation limit.

How to follow the rollout

Ultra Yearly subscribers can check their Kling account for Flash early access. For the full model, HDR and extension, follow the official release notes and the options available in the creation page. The announcement summarized here does not establish a public API launch date or third-party integration availability.