Kling AI 2.0 lands with direct video element editing

Anyone who has tried creating videos with AI knows the pain of rolling the dice. You finally get a character looking right, but a weird stray artifact ruins the background, and changing a single prompt prompt resets everything into a totally different scene.

Kuaishou wants to make the editing process behave more like actual production. At its Model 2.0 event in Beijing on April 15, the company officially launched Kling 2.0 along with Kolors 2.0, its companion image generation model.

Kling has been moving fast. Since its debut last June, the tool has rolled out more than 20 iterations and crossed 22 million global users. According to Kuaishou Senior VP Gai Kun, over 15,000 developers have plugged into Kling's API, generating roughly 12 million images and over 40 million videos for enterprise partners like Xiaomi, AWS, Alibaba Cloud, and Freepik.

Real editing replaces the slot machine

Instead of making users regenerate an entire shot, Kling 2.0 introduces Multi-modal Visual Language (MVL) and a brand new Multi-Elements Editor.

The feature allows creators to take an existing video clip and swap, insert, or remove specific elements using text prompts or reference images. Kling AI head Zhang Di explained that by combining reference frames and video clips, users can dial in identities, lighting, expressions, and camera motions without breaking the rest of the clip.

Zhang also noted an interesting stat: about 85 percent of Kling's video generation comes from image-to-video workflows. When the starting still frame carries that much weight, the underlying image engine matters immensely.

To shore up that foundation, Kolors 2.0 received a major overhaul. The update brings improved cinematic fidelity, tighter prompt adherence, and native support for more than 60 distinct visual styles.

To lock in creative talent, Kuaishou also announced the Kling AI NextGen Initiative, offering millions in funding, branding support, and early feature access to AI filmmakers.

Why it matters

We think Kuaishou is targeting the right bottleneck. Rolling a 5-second cinematic clip is easy now, but turning it into a coherent story has been virtually impossible because users lacked surgical control over the frame.

Focusing heavily on the image-to-video pipeline makes total sense. Creators usually spend hours perfecting a hero character in a 2D generator before bringing it to life. By pairing an upgraded Kolors 2.0 directly with precision video editing, Kling is trying to become an end-to-end studio rather than a quick demo site.

The test will be consistency under heavy motion. While swapping an object in a static shot is relatively straightforward, preserving lighting consistency and physics across dynamic camera angles is notoriously difficult. If Kling 2.0 pulls it off reliably, it will set a tough bar for competitors like Runway, Pika, and Google Veo to match.