Meta Researcher 남길주: Generative AI Makes 3D Modeling Unnecessary for Video Production, Graphics Shifts to Robotics
Meta Reality Labs researcher 남길주 spoke at AI Tech Day 2026 at Korea University, using his own experience of roughly five years researching 3D hair reconstruction to illustrate how generative AI can now produce realistic video without building complex 3D models. He argued that 3D graphics expertise is shifting from film and game production toward robotics and autonomous driving, domains that require spatial and physical reasoning.
ImportanceMaterialEvidenceE3 inspectableWrite-upStandard
Generative AI is making 3D modeling unnecessary in video production. In a talk at AI Tech Day 2026 held at Korea University, Meta Reality Labs researcher 남길주 offered a specific judgment: when the final output is an image or video, there is no longer a need to build a 3D model, because generative AI can already produce realistic visuals directly from text prompts. This means the center of gravity in 3D graphics is shifting from film and games toward robotics, autonomous driving and other domains that require understanding the physical world.
Traditional computer graphics works by first constructing the 3D shape of characters and objects, then computing movement and light reflections to render the scene. Generative AI skips these steps — it learns from vast amounts of image and video data and produces new visuals directly from text descriptions. Early AI video mainly generated faces; later it learned to handle complex backgrounds and light reflections. OpenAI's Sora can already realistically reproduce human gait and surrounding environments, and recently techniques have emerged that control a character's facial expressions and body movements.
남길주's own experience provides practical evidence for this judgment. He has been researching computer vision and 3D graphics at Meta Reality Labs since 2020, and spent roughly five years working on 3D hair reconstruction. He eventually changed his approach: instead of hand-designing complex graphics pipelines, he applied only basic knowledge and let AI learn the rest from data. He described the result as better than the original method.