Kuaishou's Kling AI Unveils Kling 4.0 with 30-Second Video
Kuaishou's Kling AI has introduced its latest video model, Kling 4.0, featuring capabilities for single-pass clips extending up to 30 seconds. This new iteration supports up to 15 multimodal references and allows for the setting of 10 keyframes. A lighter version, Kling 4.0 Flash, became available to Ultra Yearly subscribers last month, with the full model scheduled for release this month.
According to Pandaily, Kling 4.0 Flash, a more streamlined variant, was made accessible to a restricted group on the same day as the initial announcement. This Flash version specifically targets Ultra Yearly subscribers. Kling has not yet disclosed an exact launch date for the complete model, only indicating its arrival within this month.
The most notable alteration in Kling 4.0 involves the duration of generated video. The system now produces up to 30 seconds of video in a single pass, which is twice the 15-second maximum of the preceding Kling 3.0 generation. This improvement supports the creation of extended takes and continuous narrative within a single video segment. A forthcoming video extension feature is also planned, which will permit users to prolong a clip multiple times, potentially reaching approximately two minutes in total length.
Control features represent another significant area of development within Kling 4.0. A single task can now integrate as many as 15 multimodal references. These references may be sourced from up to 10 images, five video clips, and seven distinct subjects. This allows for closer adherence of characters, motion, camera work, and audio to the provided input material.
And users possess the capability to establish up to 10 keyframes. These keyframes serve to precisely define character states, mark scene transitions, and articulate specific plot developments. Furthermore, the prompt input capacity has been increased to 8,000 tokens, enabling the model to expand upon and refine more concise prompts.
Editing tools within Kling 4.0 have also been enhanced. Users can add, modify, or remove subjects and backgrounds from existing video content. These tools further permit adjustments to style, weather conditions, colour palettes, material textures, shot sizes, or camera viewpoints. This range of functions provides extensive post-generation refinement options.
A new remake mode has been incorporated, designed to analyse the shot language and pacing of a reference video. From this analysis, the system constructs a new commercial clip that emulates the structural characteristics of the original. This particular mode is aimed at producing content suitable for various platforms, including social media, product advertisements, fashion and beauty campaigns, and consumer electronics promotions.
On the output side, Kling 4.0 supports 4K and 1080p resolutions, presented in 10-bit HDR format, though this specific feature is listed as pending release. The system will also offer a 21:9 ultra-wide format, complete with dual-channel stereo audio and more precise lip synchronisation. The organisation states that fast motion, tracking shots, and orbiting camera movements exhibit improved stability within the new model.
Speech generation capabilities extend to multiple languages, covering Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and Hindi. This includes various accents such as American, British, and Indian English, alongside several Chinese dialects, including Cantonese. The model is also designed to render multilingual on-screen text, emoji, and corporate logos.
So, Kling 4.0 Flash specifically addresses the requirements of high-volume content creation by offering faster generation speeds at a reduced operational cost. The creation page has undergone a redesign, consolidating image, video, and audio references into a single input box. This updated interface also includes a timeline for previewing and editing, along with a canvas workspace equipped with an agent assistant. Public information regarding the API pricing for this new model remains unannounced.
Kling 4.0 generates single-pass video clips up to 30 seconds.
The model accepts up to 15 multimodal references, including images, video clips, and subjects.
Users can set 10 keyframes for precise control over character states and scene changes.
Enhanced editing tools allow for extensive customisation of existing video content.
Kling 4.0 Flash offers faster generation and lower cost for high-volume creation.
Source: Pandaily
