Kling 4.0 adds stereo audio and a faster Flash model

More control over a generation

Kling AI says the updated Omni Reference can take up to 15 reference materials in a single generation: up to ten images, five videos with a combined length of no more than 30 seconds, and seven elements. References can include storyboards, interface wireframes, image grids, videos with white models and clips with depth maps.

Creative Recreation uses an existing video as a template, allowing users to replace products or characters. Users can also change a facial expression, body movement, camera angle or background without altering the rest of the clip.

The quality comparison is still open

The comparison with Chinese competitors Wan 3.0 and Seedance 2.5 finds Kling 4.0 broadly comparable, but not ahead — particularly on shot transitions, physics and natural-sounding dialogue. That is a less decisive result than the feature list might suggest.

The comparison may be an early sign that rapid progress in video models is slowing. The next releases from Wan and Seedance will make that clearer.

Sound, language and speed

Kling 4.0 adds two-channel stereo audio and improved lip sync for more than ten languages, including German, English, Japanese, Korean and Spanish, as well as different dialects and accents. Text rendered inside videos now works across multiple languages and remains stable as the camera moves.

10-bit HDR output at 4K and 1080p, along with the ability to extend videos to two minutes, are listed as future additions. API pricing has not been announced; support for Kling 4.0 is planned for a later date.

I think the release makes a stronger case for Kling as a more controllable production tool than as a model that has pulled ahead on output quality. The missing detail is the API price: without it, the promise of cheaper generation remains a product claim rather than a cost advantage developers can assess.