Kuaishou's Kling AI Announces Version 4.0 With 8000-Token Prompts and Stereo Lip-Sync

THE ESSENTIALS

Kling 4.0 expands prompt input capacity to 8000 tokens and supports nine languages including Chinese, English, Japanese and Korean.
The model now generates up to 30 seconds of native video in a single pass, with multi-keyframe input for precise narrative control.
New stereo audio support includes improved lip-sync accuracy, targeting commercial-grade video production.
Kling 4.0 Flash is currently available to Black Gold annual members, with full rollout scheduled for October.

Major Capability Upgrade

Kuaishou's Kling AI announced on September 28 that version 4.0 will fully launch in October, marking a significant upgrade to one of China's leading AI video generation models. A Flash version is already available to Black Gold annual card members for early access.

The upgrade targets three core dimensions: visual realism, creative controllability, and narrative completeness. According to Kuaishou, Kling 4.0 delivers substantial improvements in handling high-speed motion, continuous action sequences, tracking shots, and orbital camera movements with greater stability and coherence.

Expanded Creative Controls

Kling 4.0 introduces support for up to 15 multimodal reference inputs per generation, including 10 images, five video clips, and seven subject combinations. The system can analyze reference video shot language and narrative structure to generate new commercial-grade content.

The model supports single-pass generation of up to 30 seconds of native video, addressing long-shot and continuous narrative needs. Creators can use up to 10 multi-keyframe inputs to precisely control character states and plot progression. An extended continuation feature supporting up to two minutes of video is expected to launch soon.

Audio and Multilingual Support

The new version adds support for high-quality audio and dual-channel stereo sound, with significantly improved lip-sync matching accuracy. Output formats include 21:9 ultra-wide aspect ratio, 4K resolution, and upcoming 1080p 10-bit HDR.

Prompt input capacity has been expanded to 8000 tokens, with support for Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, and Hindi, along with multiple accents and dialects.