SAM 3D Objects

  • SAM 3D

byMeta

Convert single photographs into production-ready textured 3D meshes with zero-shot semantic segmentation and UV extraction

Reconstruct 3D Mesh
SAM 3D Objects

How SAM 3D Objects works

Go from a single static photograph to an editable, textured 3D mesh in three straightforward steps.

Upload a single photograph

Upload a single photograph

Provide a clear, well-lit photo of any object, from e-commerce catalog products to field snapshots.

Target the subject

Target the subject

Specify an optional text prompt and tune the detection threshold to isolate items in busy scenes.

Export textured GLB

Export textured GLB

Download your 3D asset as a lightweight vertex-color model or a fully baked UV textured mesh.

What SAM 3D Objects is good at

Zero-shot segmentation, intelligent occlusion completion, and automated UV texture extraction combine to deliver production-ready 3D assets from single images.

Single-View Neural Reconstruction

Single-View Neural Reconstruction

Reconstruct complete 360-degree 3D geometry from a single 2D photograph by inferring unseen sides and back surfaces through deep conceptual priors.

Prompt-Guided Segmentation

Prompt-Guided Segmentation

Isolate individual items from cluttered environments using short text prompts and adjustable confidence thresholds ranging from 0.1 to 1.0.

Baked UV Texture Export

Baked UV Texture Export

Toggle export of textured GLB meshes with baked UV coordinates for direct import into game engines, AR viewers, and standard DCC applications.

Photographic Surface Fidelity

Photographic Surface Fidelity

Preserve organic surface details including leather grain, fabric weave, and ceramic glaze directly projected from the source image.

Made with SAM 3D Objects

Explore real-world objects digitized from single photos into textured 3D meshes with organic geometries and realistic surface qualities.

Anodized aluminum and cork furniture mesh

Anodized aluminum and cork furniture mesh

Precision optic hardware with textured mechanical details

Precision optic hardware with textured mechanical details

Glazed ceramic sanitaryware with fine surface fluting

Glazed ceramic sanitaryware with fine surface fluting

Translucent footwear prototype isolated from editorial photo

Translucent footwear prototype isolated from editorial photo

What people build with SAM 3D Objects

From e-commerce merchandising to game development and spatial design, creators reach for SAM 3D Objects to accelerate 3D asset workflows.

E-Commerce Product Digitization

01

Convert catalog photos into interactive 3D assets for web viewers, capturing real fabric weaves, leather grain, and hardware finishes without multi-camera studio rigs.

Spatial Asset Prototyping

02

Generate dimensionally consistent GLB props from concept sketches or reference photos to quickly populate mixed-reality scenes and spatial computing environments.

Game Asset Creation from Field Photos

03

Isolate and reconstruct real-world artifacts, street furniture, and architectural fragments from snapshot photography directly into game-ready 3D meshes.

Digital Apparel and Footwear Archiving

04

Transform runway photography and lookbook stills into three-dimensional wearable models with baked UV textures for virtual showrooms.

Virtual Interior Staging

05

Extract bespoke furniture, handcrafted ceramics, and vintage decor items from interior photos to populate 3D architectural visualizations.

Frequently Asked Questions

Yes, SAM 3D is the common abbreviation for SAM 3D Objects, Meta's model for single-image 3D object reconstruction. It builds upon Segment Anything Model 3 (SAM 3) segmentation to isolate target objects and reconstruct complete 3D GLB meshes from a single 2D image.

Use high-resolution photos (1024px to 2048px on the shortest dimension) taken under soft, diffuse studio lighting. The target object should be fully visible without edge clipping, ideally captured from a three-quarter or front-facing angle. Soft lighting prevents harsh shadows from baking directly into the final 3D texture.

Avoid highly reflective metals, polished chrome, mirrors, and transparent glass, as optical refractions disrupt the neural depth estimation pipeline. Additionally, steer clear of 2D stylized illustrations and heavily occluded objects where more than 40% to 50% of the item is hidden from view.

Enabling the Export Textured GLB option generates a 3D model with baked UV coordinate texture maps, making it directly importable into game engines and web 3D viewers. Disabling the toggle outputs a lightweight mesh relying strictly on vertex colors for stylized or high-poly geometric workflows.

Enter an optional text prompt describing the target object (such as leather boots or wooden chair) and adjust the Detection Threshold between 0.1 and 1.0. Lowering the threshold helps capture faint or low-contrast objects, while raising it prevents surrounding background clutter from fusing into the mesh.

SAM 3D Objects was developed by Meta, leveraging the Segment Anything Model 3 foundation. It was released following the SAM 3 suite between late 2025 and early 2026.

Try SAM 3D Objects on Fuser

Convert single photographs into production-ready textured 3D meshes with zero-shot semantic segmentation and UV extraction

Reconstruct 3D Mesh