Jiseob Kim

Kao Corporation (Japan), Seoul National University

Papers

Total Citations

146

H-Index

About

Jiseob Kim is a researcher at the forefront of computer vision and human motion synthesis, with a particular focus on bridging the gap between natural language and physical movement. His most impactful contribution is the development of **FLAME (Free-Form Language-Based Motion Synthesis & Editing)**, a diffusion-based model that enables text-driven generation and editing of 3D human motions. This work, published in 2023, has already garnered **121 citations**, reflecting its significant influence on the game, animation, and robotics industries by automating complex motion-making processes. Earlier in his career, Kim explored human action recognition, proposing a two-stage framework that combined unsupervised spatio-temporal feature learning (via Independent Subspace Analysis) with semantic rules to automatically extract activities from video. This foundational work, cited **21 times**, demonstrated his early commitment to making machines understand human behavior. Kim’s research trajectory—from action recognition to cutting-edge language-based motion control—showcases his ability to evolve with the field, making him a notable figure in the growing intersection of generative AI and embodied intelligence.