Chinese short-video app Kuaishou has taken a major step in the generative AI race by launching its own text-to-video service called the Kling AI Model. This new offering is seen as a direct competitor to OpenAI’s Sora, as Chinese tech firms scramble to catch up with their US counterparts in the rapidly evolving field of artificial intelligence applications. The Kling AI Model, currently in trial stage, can process text prompts to generate video clips up to 2 minutes long with 1080p high-definition resolution. It supports various aspect ratios and can interpret instructions to create realistic videos mimicking the physical world or imaginative scenes from text descriptions.
Kuaishou has released several demonstration videos showcasing Kling’s capabilities, including a white cat driving a car through busy city streets, a boy eating a cheeseburger in a fast-food restaurant, and a child riding a bicycle as the seasons change around them. With the text-to-video feature, Kuaishou joins the ranks of several Chinese tech companies racing to launch challengers to OpenAI’s Sora. In April, local start-up Shengshu Technology, in collaboration with Tsinghua University, launched a similar tool called Vidu, while generative AI firm Zhipu AI is also reportedly working on its own text-to-video model.
Kling AI is Kuaishou’s latest foray into AI innovations, following the release of its KwaiYii large language model and the text-to-image model Kolors in May. The company has also introduced an “AI Dancer” feature that generates videos of characters dancing to specific rhythms from still photos, and plans to launch an image-to-video capability based on Kling soon. According to Kuaishou, Kling AI uses a 3D space-time attention system to accurately model motion and physical interactions between objects. It also employs a diffusion transformer to combine concepts and create fictional scenes, allowing the model to generate long, high-resolution videos efficiently.
While both Kling and Sora utilize diffusion transformers, Kling can produce longer videos of up to two minutes at higher 1080p resolution compared to Sora’s one-minute limit. The launch of Kling AI is part of Kuaishou’s broader push into AI as it aims to capitalize on the generative AI boom. The company recently reported strong quarterly results, with a net profit of $575.1 million driven largely by its online marketing services and e-commerce business.
As generative AI for video continues to advance, tech investor Ashton Kutcher believes it could revolutionize the film market and Hollywood, transforming content creation and production in the entertainment industry.
