Dexter Studios(KOSDAQ: 206560), a leading content technology company, announced that it has been selected as the lead organization for the national research project titled ‘Development of Compression, Restoration, and Rendering Technologies for Free-Viewpoint Media’, sponsored by the Ministry of Science and ICT and the Institute for Information & Communications Technology Planning & Evaluation (IITP). The consortium brings together top-tier research institutions, including Seoul National University, Yonsei University, POSTECH, Sungkyunkwan University, and Korean Broadcasting System(KBS).
The core objective of this project is to develop key technologies for high-efficiency compression, restoration, scene editing, and real-time rendering of 4D Gaussian Splatting-based free-viewpoint media, while establishing domestic industry standards.
3D Gaussian Splatting is a technology that leverages AI to render a series of captured photographs into realistic 3D environments in real time. By training on just a few images, AI infers uncaptured angles to build realistic 3D spaces without traditional 3D modeling. 4D Gaussian Splatting adds a time axis to this technology, allowing AI to render dynamic scenes—including object movements, deformations, and physical dynamics—at a photorealistic level in real time. This enables viewers to freely adjust their viewing angles in a moving scene, as if moving a camera 360 degrees around the action.
Driven by breakthrough rendering speeds and high scene fidelity, 3D and 4D Gaussian Splatting technologies are being actively researched and applied by global tech giants such as Google, Meta, and NVIDIA across diverse domains, including XR content creation, spatial scanning, digital twins, and physical AI.
However, Gaussian Splatting requires storing millions of Gaussian primitives, resulting in massive data volumes. Its closed formats also pose challenges for real-time streaming on multi-access platforms like mobile and XR devices. To address these issues, Dexter Studios aims to establish technology sovereignty for the K-content industry rather than relying on foreign solutions, while developing high-efficiency compression, restoration, and rendering standards to maximize production efficiency.
The research will focus on acquiring high-quality free-viewpoint media data under various indoor and outdoor static/dynamic conditions, including sparse viewpoints, and training a high-efficiency 4D Gaussian generation model tailored for dynamic scenes.
Following data acquisition, the team will build a customized compression pipeline where AI independently evaluates data importance to lightweight the output. This framework will then be integrated into commercial VFX production pipelines, verifying production efficiency across feature films, OTT content, and broadcast programs.
Integrating 3D and 4D Gaussian Splatting into VFX production pipelines is expected to streamline processes such as background, object, and crowd compositing, as well as free-viewpoint camera direction.
Jaewon Song, Head of Dexter’s R&D Center, stated: “3D and 4D Gaussian Splatting technologies dramatically reduce production time and costs while delivering high-quality 3D spatial data. Beyond VFX and XR content creation, these technologies are drawing attention across various industries, including digital twins and physical AI. By assetizing free-viewpoint media through libraries or sharing platforms, we can generate new value-added business models, such as commercial licensing.”
▲Dexter Studios CI
