Graphics Research News
The latest Graphics Research coverage curated for game developers.
arXiv cs.GR details GS-PI, a method for turning Gaussian-splat assets into fully relightable PBR content without per-scene joint optimization. The approach separates appearance decomposition from geometry, then uses diffusion and …
A SIGGRAPH 2026 paper on dichoptic foveation suggests VR renderers can push more perceived detail without fully sharpening both eyes. The technique blurs one eye while sharpening the other, then …
HairCS turns hair-card assets into strand-based hairstyles, aiming to keep the original silhouette while adding finer geometric detail. The pipeline also enforces production-friendly constraints like scalp-rooted strands and even root …
PhysStream pushes video generation toward physically meaningful control, letting creators steer motion mid-generation instead of locking in a full schedule up front. The model uses structured scene memory and sparse …
Marching Cubes gets a sharp conservative offset for grid-sampled 1-Lipschitz fields, with a provable threshold of √3/2·h. For signed distance functions and similar data, that means a guaranteed enclosure of …
arXiv cs.GR details TopoRig, a facial rigging system that keeps a character’s original mesh topology while predicting FACS-driven deformations. For teams building custom characters, the pitch is fewer transfer artifacts …
A new compression method for 3D Gaussian splats targets one of the format’s biggest practical costs: color attributes. By transforming spherical harmonic coefficients before coding, the researchers report more than …
Researchers have introduced Unified Gaussian Primitives, a scene representation and rendering approach aimed at consolidating how complex content is stored and drawn. Published in ACM Transactions on Graphics, the work …
DealMaTe proposes a diffusion-transformer approach for multi-dimensional material transfer, aiming to move more than just color or texture between assets. Published in ACM Transactions on Graphics, the work points toward …
Researchers have proposed CVQPG, a hologram representation that swaps the usual 2D Gaussian basis in splatting-style pipelines for quadratic phase functions. In tests, it improved reconstruction quality by 0.19 dB …
A new Gaussian splatting pipeline swaps bulky spherical harmonics for compact neural appearance codes, cutting per-primitive appearance storage from 192 bytes to 28. The result is a 1.3× faster optimization …
TileGS is a new tile-local depth binning approach for 3D Gaussian Splatting that cuts raster traversal work without changing output quality. On a 9-scene benchmark, the No-GW variant sped up …
A new 3D Gaussian Splatting editing method lets artists recolor scenes in real time with palette-based controls and separate luminance adjustment. The technique reparameterizes pretrained 3DGS instead of retraining from …
A new self-supervised method called P-CORE aims to make point-based neural editing survive large non-rigid deformations without ground-truth views of the edited shape. It enforces surface consistency before and after …
LightBridge aims to make relighting 3D Gaussian Splatting assets a one-pass process instead of a per-scene optimization job. The system uses a feed-forward generative pipeline plus a new relighting dataset …
Researchers have introduced an inverse rendering method that reconstructs fuzzy geometry with explicit line segments instead of volumetric primitives. The approach targets hair, fur, fibers, and textiles, aiming for real-time-friendly …
Researchers have introduced an interactive way to deform 3D Gaussian Splatting scenes while keeping the rendered result consistent with the edited shape. The method targets a long-standing problem in 3DGS …
mmIR tackles a core radar bottleneck for simulation-heavy development: there still isn’t much high-resolution 3D millimeter-wave data to train or validate against. The system fits a physics-based forward model to …
A new training method for 3D Gaussian Splatting cuts peak VRAM to a fixed budget by training radiance fields block by block. ABCD pre-renders frozen regions into RGBA composites, letting …
A new image-editing framework, InteractEdit, tackles human-object interaction changes without needing a reference image. It aims to preserve subject and object identity while swapping the interaction itself, a harder problem …