Sharp Monocular View Synthesis in Less Than a Second

2025/12/11 by Lars Mescheder, Wei Dong, Mescheder, Lars +24 · 2 voices · 2 citations
Computer Science · Engineering · #3D Shape Modeling and Analysis #Advanced Vision and Imaging #Generative Adversarial Networks and Image Synthesis #cs.CV #cs.LG

paper · pdf · doi:10.48550/arxiv.2512.10685

openalex publication_date 2025/12/11 · openalex created_date 2025/12/13 · openalex updated_date 2026/07/28

Abstract

We present SHARP, an approach to photorealistic view synthesis from a single image. Given a single photograph, SHARP regresses the parameters of a 3D Gaussian representation of the depicted scene. This is done in less than a second on a standard GPU via a single feedforward pass through a neural network. The 3D Gaussian representation produced by SHARP can then be rendered in real time, yielding high-resolution photorealistic images for nearby views. The representation is metric, with absolute scale, supporting metric camera movements. Experimental results demonstrate that SHARP delivers robust zero-shot generalization across datasets. It sets a new state of the art on multiple datasets, reducing LPIPS by 25-34% and DISTS by 21-43% versus the best prior model, while lowering the synthesis time by three orders of magnitude. Code and weights are provided at https://github.com/apple/ml-sharp

Citations

• A Recipe for Generating 3D Worlds From a Single Image
• Stable Virtual Camera: Generative View Synthesis with Diffusion Models
• Bolt3D: Generating 3D Scenes in Seconds
• Seeing and Seeing Through the Glass: Real and Synthetic Data for Multi-Layer Depth Estimation
• GEN3C: 3D-Informed World-Consistent Video Generation with Precise Camera Control
• High-Fidelity Novel View Synthesis via Splatting-Guided Diffusion
• Wonderland: Navigating 3D Scenes from a Single Image
• You See it, You Got it: Learning 3D Creation on Pose-Free Videos at Scale
• MoGe: Unlocking Accurate Monocular Geometry Estimation for Open-Domain Images with Optimal Training Supervision
• LVSM: A Large View Synthesis Model with Minimal 3D Inductive Bias
• Depth Pro: Sharp Monocular Metric Depth in Less Than a Second
• ViewCrafter: Taming Video Diffusion Models for High-fidelity Novel View Synthesis
• MegaScenes: Scene-Level View Synthesis at Scale
• WonderWorld: Interactive 3D Scene Generation from a Single Image
• Flash3D: Feed-Forward Generalisable 3D Scene Reconstruction from a Single Image
• NeRF-Casting: Improved View-Dependent Appearance with Consistent Reflections
• CAT3D: Create Anything in 3D with Multi-View Diffusion Models
• GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting
• UniDepth: Universal Monocular Metric Depth Estimation
• RGBD Objects in the Wild: Scaling Real-World 3D Object Learning from RGB-D Videos
• DUSt3R: Geometric 3D Vision Made Easy
• Splatter Image: Ultra-Fast Single-View 3D Reconstruction
• One-step Diffusion with Distribution Matching Distillation
• LRM: Large Reconstruction Model for Single Image to 3D
• ZeroNVS: Zero-Shot 360-Degree View Synthesis from a Single Image
• iNVS: Repurposing Diffusion Inpainters for Novel View Synthesis
• State of the Art on Diffusion Models for Visual Computing
• State of the Art on Diffusion Models for Visual Computing
• Tiled Multiplane Images for Practical 3D Photography
• ScanNet++: A High-Fidelity Dataset of 3D Indoor Scenes
• 3D Gaussian Splatting for Real-Time Radiance Field Rendering
• Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fields
• Zero-1-to-3: Zero-shot One Image to 3D Object
• NerfDiff: Single-image View Synthesis with NeRF-guided Distillation from 3D-aware Diffusion
• Booster: a Benchmark for Depth from Images of Specular and Transparent Surfaces
• Novel View Synthesis with Diffusion Models
• Single-View View Synthesis in the Wild with Learned Adaptive Multiplane Images
• FILM: Frame Interpolation for Large Motion
• Resolution-robust Large Mask Inpainting with Fourier Convolutions
• SLIDE: Single Image 3D Photography with Soft Layering and Depth-aware Inpainting
• Advances in Neural Rendering
• MVSNeRF: Fast Generalizable Radiance Field Reconstruction from Multi-View Stereo
• Vision Transformers for Dense Prediction
• IBRNet: Learning Multi-View Image-Based Rendering
• pixelNeRF: Neural Radiance Fields from One or Few Images
• Stable View Synthesis
• An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
• Free View Synthesis
• On the uncertainty of self-supervised monocular depth estimation
• Single-View View Synthesis with Multiplane Images
• SynSin: End-to-end View Synthesis from a Single Image
• DeepView: View Synthesis with Learned Gradient Descent
• The Unreasonable Effectiveness of Deep Features as a Perceptual Metric
• Unsupervised Monocular Depth Estimation with Left-Right Consistency
• SGDR: Stochastic Gradient Descent with Warm Restarts
• View Synthesis by Appearance Flow
• Perceptual Losses for Real-Time Style Transfer and Super-Resolution
• U-Net: Convolutional Networks for Biomedical Image Segmentation

Cited by

• Genie Sim PanoWorld: An Infinite Indoor 3D World Generation Pipeline via Panoramic Scene Modeling and Simulation
• Compact Representation of Mipmapped SVBRDFs via Shared Gaussians

Discussions

• Sharp Monocular View Synthesis in Less Than a Second Lars Mescheder, Wei Dong, Shiwei Li, Xuyang Bai, Marcel Santos, Peiyun Hu, Bruno Lecouat, Mingmin Zhen, Amaël Delaunoy, Tian Fang, Yanghai Tsin, St [bsky, 5 points, 0 comments]
• Apple's SHARP model generates photorealistic 3D Gaussian reps from a single image. Free and Open source project. GitHub: github.com/apple/ml-sharp Paper: arxiv.org/abs/2512.10685 [bsky, 1 points, 0 comments]

Related

• Fast and Explicit Neural View Synthesis
• FWD: Real-time Novel View Synthesis with Forward Warping and Depth
• D-NPC: Dynamic Neural Point Clouds for Non-Rigid View Synthesis from Monocular Video
• One View Is Enough! Monocular Training for In-the-Wild Novel View Generation
• Efficient-3DiM: Learning a Generalizable Single-image Novel-view Synthesizer in One Day