The article explains how InSpatio's QuerySplat and Topos‑Lite models replace pixel‑aligned 3D Gaussian Splatting with a query‑based Transformer approach, allowing a handful of uncalibrated photos to be turned into high‑quality, explorable 3D scenes in seconds while achieving state‑of‑the‑art benchmark results.
TryOnCrafter introduces a camera‑controllable video virtual try‑on framework that builds a renderable 4D proxy to enable free‑view, 360° and bullet‑time effects while preserving structural stability, texture consistency, and realistic motion across arbitrary camera trajectories.
UniSHARP is an open‑source monocular 3D Gaussian Splatting model that, with a single input image, instantly produces a high‑quality Gaussian point cloud for perspective, wide‑angle, fisheye and 360° cameras, eliminating the need for multi‑view inputs or per‑scene optimization.
The Meta team’s CCIG 2026 presentation details the TaoAvatar system—a 3D Gaussian‑splatting based pipeline that captures multi‑view video, generates high‑quality static and dynamic digital humans, drives them with voice, text and gestures, and powers the immersive Taobao Vision shopping experience across XR devices.
The SIGGRAPH 2026 paper “Generative 3D Gaussians with Learned Density Control” introduces Density‑Sampled Gaussians (DeG), a differentiable framework that lets a model learn where to place Gaussian splats by sampling from a learned spatial density, enabling arbitrary‑budget, non‑uniform 3D representations with higher quality per cost.
GS‑Playground is a next‑generation visual‑high‑fidelity robot simulator that cuts photo‑level rendering cost, automates asset creation, and narrows the Sim2Real gap, achieving up to 10,000 FPS on RTX 4090 and outperforming MuJoCo by 32× while supporting full‑stack parallel physics, 3DGS rendering, and end‑to‑end Real2Sim pipelines.
PhysGM, a CVPR 2026 paper by Beijing Institute of Technology and Li Auto, transforms a single static image into a high‑fidelity 4D video that obeys real‑world physics in under a minute, using a dual‑decoder transformer, DPO alignment, and a newly built 50k‑item PhysAssets dataset, outperforming prior methods in speed and quality.
The Xiaomi automotive research team secured first place in the ICCV 2025 RealADSim New View Synthesis challenge by presenting a four‑stage, 3D Gaussian Splatting‑based pipeline that tackles LiDAR‑free initialization, structured road modeling, generative pseudo‑ground‑truth refinement, and time‑invariant adaptation, achieving top PSNR, SSIM, and LPIPS scores.
