Chinese technology company Kuaishou Technology, known for its content community and social platform, unveiled Kling AI, a new tool designed to generate highly realistic videos from text prompts, representing a potential competitor to OpenAI’s Sora model.
Sora is an AI model designed to generate lifelike and creative scenes based on textual input. Central to Sora’s video generation capabilities is a deep learning framework. The model learns associations between words and phrases with various elements such as objects, actions, environments, and stylistic nuances.
While OpenAI’s Sora model is capable of generating videos up to one minute in length, the Kling AI can produce realistic videos reaching up to two minutes with 1080p resolution at 30 frames per second. It supports multiple aspect ratios and achieves a high level of detail through advanced 3D face and body reconstruction techniques, ensuring that its AI-generated videos are virtually indistinguishable from real footage.
Kling AI utilizes a 3D Variational Autoencoder (VAE) to reconstruct faces and bodies, enabling the generation of detailed expressions and limb movements from a single full-body image. This technology is enhanced by a 3D spatiotemporal joint attention mechanism, enhancing the model’s ability to handle complex scenes and movements. This ensures that the generated content maintains consistency with the laws of physics.
The application adopts an open-access approach, although with regional limitations in place. The model is currently functioning in its trial phase.
Less than 48 hours ago, Sora competitor Kling dropped.
People are already getting access and creating wild AI videos. 
1. MadMax Beer commercial made in 1 hourpic.twitter.com/CyKm2aI0It
— Min Choi (@minchoi) June 8, 2024
Kuaishou Technology Unveils KwaiYii LLM And Kolors Text-to-Image Model
Kuaishou Technology is among several Chinese technology companies competing to introduce alternatives to Sora. The text-to-video capability represents its most recent AI innovation, following the introduction of its KwaiYii large language model (LLM) and the text-to-image model Kolors, which debuted in May.
Additionally, the “AI Dancer” feature, accessible within the Kuaishou Technology application and the video production application Kwaiying, enables users to upload a static photo and generate videos featuring the character dancing to specified rhythms and movements.
The post Kuaishou Technology Unveils Kling AI, Empowering Users To Create Videos Via Text-to-Video Generation appeared first on Metaverse Post.