Sihyun Yu

Title: Sihyun Yu and High-Resolution Video Generation

Introduction

Sihyun Yu is an inventor based in Gyeonggi-do, South Korea, and works at NVIDIA Corporation. His patent work includes a system and method for generating high-resolution videos from text prompts.

Latest Patents

Yu is associated with 1 patent: “System and Method for Efficient Text-Guided Generation of High-Resolution Videos.” The invention describes training and using a content frame-motion latent diffusion model to generate requested videos.

The disclosed approach uses a two-stage framework. First, an autoencoder learns a compact latent representation that separates video content frames from motion. Then, a pretrained image diffusion model fine-tunes the content-frame distribution, drawing on its learned visual knowledge. A lightweight diffusion model generates motion representations conditioned on the content frame, enabling the system to produce videos from text guidance.

Career Highlights

Yu’s patent reflects work at the intersection of generative artificial intelligence, video modeling, and diffusion-based methods. The approach combines image-generation capabilities with a separate representation of motion to support high-resolution video synthesis.

Collaborations

The patent information names Weili Nie and De-An Huang as Yu’s coworkers. Their work is associated with the disclosed video-generation system.

Conclusion

Sihyun Yu’s patent describes a two-stage diffusion framework that combines learned video content and motion representations to generate high-resolution videos from text prompts.