SAM 3D

SAM 3D is a platform by Meta AI that converts single images into high-fidelity 3D models for objects and human bodies.

What is SAM 3D?

SAM 3D is a web-based platform developed by Meta AI (based on their open-source SAM 3D models) that reconstructs detailed 3D meshes, textures, and poses from a single RGB image. It supports two specialized models: SAM 3D Objects for general objects and scenes, and SAM 3D Body for human pose and shape estimation. The platform requires no setup or GPU, runs entirely in the browser, and exports to standard 3D formats like .OBJ and .GLB. Try SAM 3D Now or Try SAM 3 Now for image cutout.

Key Features

• Open Vocabulary Support — Reconstruct any object category without limitation, thanks to large-scale pre-training on diverse concepts.
• Meta Momentum Human Rig (MHR) — A new open-source mesh format separating skeletal structure from soft tissue for superior animation.
• Occlusion Handling — Advanced inference logic hallucinates plausible back-facing geometry for occluded objects.
• Multi-View Consistency — Ensures texture and geometric consistency across different viewing angles of the reconstructed object.
• Apache 2.0 License — Fully open-source code and model weights, enabling commercial usage and community innovation.
• SA-3DAO Dataset — Trained on 1M+ visually grounded real-world images with human-in-the-loop annotation.

Who is it for?

• E-commerce teams — Create 'View in Room' experiences by reconstructing any product from a single photo.
• Game developers — Generate 3D assets for virtual worlds from reference images.
• VR/AR creators — Digitize objects and people for immersive experiences.
• Sports analysts — Use SAM 3D Body for pose estimation and movement tracking.

What can you do with SAM 3D?

• Indie 3D artists: Quickly generate base meshes from images to speed up modeling workflows.
• Online retailers: Automatically create 3D product views from existing catalog photos.
• Fitness apps: Extract 3D human poses for exercise form analysis.

How does SAM 3D work?

1. Upload any standard RGB image (phone photo, video frame, or generated image).
2. Click on the object or person you want to reconstruct — the model instantly generates a segmentation mask.
3. SAM 3D infers 3D geometry, texture, and pose, producing a high-quality mesh in seconds.
4. Export the asset in formats like .OBJ, .GLB, or MHR (for human bodies) for use in Blender, Unity, or Unreal Engine.

FAQ

What is the difference between SAM 3D and the original SAM?

The original SAM (Segment Anything Model) produces 2D segmentation masks only. SAM 3D extends this to 3D reconstruction, generating full 3D meshes, textures, and poses from a single image.

Can SAM 3D handle video inputs?

The platform is designed for single images. While not directly supporting video, you can extract frames from a video and process them individually.

Is the model open source for commercial use?

Yes, SAM 3D is released under the Apache 2.0 license, allowing commercial use, modification, and redistribution.

What hardware is required to run SAM 3D?

No special hardware is needed. The platform runs entirely in a web browser with cloud-based inference, requiring no local GPU.

What is the MHR format mentioned in SAM 3D Body?

MHR (Meta Momentum Human Rig) is an open-source mesh format that separates skeletal structure from soft tissue shape, enabling better animation and posing of reconstructed human models.