Kling 4.0 Hands-On: 30-Second Native Clips Let One Person Remake a Wong Kar-wai Short

葬AI · wechat · 2026-10-19

The author had dismissed AI video as low-brow entertainment until two signals changed his mind: Korean director Kim Sung-su released an AI-made film, and Netflix admitted at Busan International Film Festival that AI was used across roughly 300 titles for previz, art design, scene design, editing and VFX. He then stress-tested Kling 4.0, which is positioned squarely at filmmaking.

Test cases

A multilingual music video mixing English, Japanese, Russian, Shanghainese and Henanese singing: lip-sync and subtitles held up under complex camera moves and similar-match cuts, with smooth transitions.
A pure dialogue piece ("I wanna go to China"): natural facial expression shifts and tone changes, using silence to convey emotion.
A fight scene using the "flash-step" meme: a 15-second single take with camera movement but no cuts, strong spatial coherence, convincing impact and bookshelf collisions, evoking classic Hong Kong action.
Stylized remakes of the viral "bushao xin" short in Wong Kar-wai, Wes Anderson and Quentin Tarantino styles: 30-second pieces stayed consistent, with BGM syncing to the plot.
A remake of a famous "Seoul Spring" scene: prompts reverse-engineered with ChatGPT, regenerated as one continuous shot, with facial muscle micro-movements driving light and shadow.

Takeaway: Kling 4.0's updates are tightly focused on filmmaking — large motions, complex camera work, up to 30-second native generation, up to 15 reference items and 10 keyframes, and strong character consistency. The downside: prompts must specify every line and action or output looks like a slideshow. The author expects future films to blend live-action and AI-generated footage indistinguishably.

Original post →

Kling 4.0 Hands-On: 30-Second Native Clips Let One Person Remake a Wong Kar-wai Short

More from Multimodal