The technology can convert text prompts into video clips and uses a similar technical route to Sora. To try out the technology, you'll need to download the Kuaishou app and join the waiting list. The public demo is currently only available in China.
A rabbit reading a newspaper. (Image courtesy of Kling)
New Delhi: Kuaishou, the second most popular short video app in China (after Douyin, the Chinese version of TikTok), has unveiled its Kling AI model, positioned as a rival to OpenAI's Sora. The model can generate high-definition videos of up to two minutes in length at 1080p resolution and 30 frames per second. The model can generate complex motion sequences and physical interactions between objects.
Kuishou demoed several clips created with Kling, including a panda playing a guitar, a cat driving a car, and a rabbit reading the newspaper in a cafe. The demo also included a boy riding his bike through a garden as the seasons change, and a knife chopping an onion. The demo was impressive, with Kling doing a good job of guessing certain species, such as an angelfish, when prompted.
Kling vs Sora
Kling uses a 3D spatio-temporal attention system to model physical interactions between objects, and a diffusion transformer to create fictional scenes that combine multiple prompted concepts. The combination of the two technologies allows Kling to consistently generate two-minute high-definition videos. While Sora only generates one-minute movies, Kling's output can be twice as long. As such, Kuaishou positions Kling as a competitor to Sora in the text-to-video conversion space.
How can I try Kling?
The public demo is only available to China residents at this time. Users must first download the Kuaishou app and register for an account, which requires a Chinese phone number to which an access code will be sent. There is a tab for AI features in the app. Within the AI tab, users must sign up for a waiting list to use Kling AI. Once access is provided, users will be able to use Kling to generate videos with text prompts.
