Ke Ling Big Model is an AI video generating big model launched by the Racer team, with powerful video creation capabilities, using 3D space-time joint attention mechanism, capable of generating a large movement video that conforms to the laws of physics, simulating real world characteristics. Can spiritually support the generation of high-definition videos up to 2 minutes in length and 1080p resolution, and has the function of freely adjusting the aspect ratio. In addition, this AI video tool also combines 3D face and human body reconstruction technology to achieve fully driven expressions and limbs, allowing you to experience vivid AI singing and dancing features with just a full-body photo.
Functional features of Kerin Big Model
- Large and reasonable motion: using 3D spatio-temporal joint attention mechanism, it is capable of modelling complex spatio-temporal motion and generating large motion videos that conform to the laws of motion.
- Up to 2-minute-long video generation: thanks to efficient training infrastructure and inference optimisation, Koring is able to generate video content up to 2 minutes long.
- Physical world simulation: Based on self-developed model architecture, we can simulate the physical properties of the real world and generate videos that conform to the laws of physics.
- Powerful Concept Combination Capability: Using a deep understanding of text-video semantics and the Diffusion Transformer architecture, we can transform users' imagination into concrete images.
- Cinema-grade picture quality generation: Based on self-developed 3D VAE technology, it can generate cinema-grade videos with 1080p resolution.
- Supports free video aspect ratio: adopts variable resolution training strategy, capable of outputting diverse video aspect ratios in the inference process.
- AI-driven innovative gameplay: Combining 3D face and human body reconstruction technology, it realises fully-driven expression limbs, and users can upload full-body photos to experience vivid AI singing and dancing gameplay.
- Graphic Video: The Kling model converts static images into 5-second dynamic videos, and users can generate diverse motion effects through text prompts.
- Video Continuation: Support one-key continuation of the existing video, each extension of 4.5 seconds, can be renewed multiple times, up to 3 minutes, to achieve the user's creativity.
