The brief

Analyst view

• Kuaishou operates one of China's largest short-video platforms with 500+ million monthly active users.
• Its AI division built Kling, a video generation model that generates videos with native 4K/60fps and synchronized audio.
• Kling 3.0 (released Feb 2026) ranked #1 on the video Elo board ahead of Veo 3.1, handling fast motion and scene changes.

Technical approach

As reported

Architecture

Diffusion transformer with 3D VAE

Diffusion transformer with 3D VAE for spatiotemporal compression and efficient full-attention motion modeling.

Video Generation

Native 4K at 60fps with audio

Generates videos at 4K (native, not upscaled) at 60fps with synchronized audio, handling fast motion and scene changes.

Input Flexibility

Text, image and video inputs

Accepts text, images and videos to guide generation for creative production workflows.