Kuaishou’s Kling AI released its 3.0 model series on February 5, with updates aimed at making AI-generated videos more consistent, easier to control and more useful for everyday creators.
One of the biggest improvements is character consistency. In earlier versions of AI video tools, the same person could look different from one shot to the next. Clothes might change, objects could disappear and simple movements sometimes looked awkward.
Kling AI 3.0 is designed to reduce those problems. Users can upload photos, product images or reference videos to help the system keep track of people, clothing and locations across several scenes.
Take a small coffee shop owner who wants to make a short promotional video. They could upload photos of the shop and barista, then write: “Early in the morning, the barista opens the café, turns on the lights, makes a fresh cup of coffee and hands it to the first customer.”
The system can then create a sequence with an outside shot of the shop, a close-up of the coffee being made and a final scene of the customer receiving the cup.
In the past, a video like this would usually require filming, editing and several pieces of equipment. Now, someone with little or no production experience can create a basic version on a laptop.
The new model can also plan multiple shots from a single description. Instead of creating each scene separately and joining them in editing software, users may be able to produce a more complete sequence in one go.
The process is still not perfect. Complicated movements can look unnatural, and vague instructions may lead to unexpected results. Most users will still need to rewrite their prompts and generate several versions before getting something they can use.
Kling AI 3.0 lowers the barrier to making short videos, especially for small businesses, independent creators and people without professional editing skills.
It can help build the shots. It can save time. But it cannot decide what is worth saying. That part still belongs to the person behind the screen.
