New Release

Create Stunning 3D Models From a Single Image

Name: SAM 3D Single Image Reconstruction - Turn Photos into 3D Models
Author: SAM 3D

SAM 3D offers a powerful one photo 3D model that reconstructs full 3D objects from a single 2D image, even with occlusions. Experience the future of single view 3D generation with Meta AI.

Generate Now Learn More

High-Quality Output

Fast Reconstruction

SAM 3D single image reconstruction example

Key Features of SAM 3D

Explore the groundbreaking capabilities that set SAM 3D apart.

Single Image 3D Reconstruction

Reconstruct complete 3D geometry, texture, and layout from a single 2D image using Meta AI's advanced algorithms. Now you can create a one photo 3D model with ease.

Gaussian Splatting for High Fidelity

Utilize Gaussian splatting representation (up to 32 splats per occupied voxel) for unparalleled, high-quality point-based neural rendering, bringing your monocular 3D reconstruction to life.

Human Mesh Recovery

SAM 3D Body model specializes in single-image 3D human mesh recovery (HMR), accurately estimating human body pose and shape, even when occlusions are present. Amazing for single view 3D generation focused on humans.

How SAM 3D Works

Transforming a single image into a full 3D model is easier than you think.

Upload Your Image

Simply upload a single 2D image of the object or scene you want to reconstruct in 3D. SAM 3D handles complex scenes and occlusions.

AI-Powered Reconstruction

Meta AI's SAM 3D algorithms automatically analyze the image and reconstruct the full 3D model, predicting shape, texture, and layout using a monocular 3D reconstruction approach.

Download Your 3D Model

Download the reconstructed 3D model in various formats, ready for use in your projects. Use this one photo 3D model to start creating today!

Frequently Asked Questions About SAM 3D

Get answers to common questions about SAM 3D single image reconstruction.

SAM 3D is a groundbreaking AI model from Meta AI designed for 3D reconstruction from a single 2D image. It generates full 3D object geometry, texture, and layout, simplifying the typically complex process of 3D modeling.

SAM 3D uses a Mixture-of-Transformers (MoT) architecture and a depth-VAE with visibility masking to accurately predict 3D structure and refine texture, even in occluded regions. Gaussian splatting ensures high-fidelity rendering.

SAM 3D achieves at least a 5:1 win rate in human preference tests over recent 3D generation models on real-world objects and scenes. It also reconstructs complete 3D objects, including occluded parts, not just visible surfaces.

SAM 3D consists of two specialized models: SAM 3D Objects for general scene reconstruction, and SAM 3D Body, which is specifically trained for accurate human mesh recovery even in challenging conditions.