SOPHY: Generating Simulation-Ready Objects with PHYsical Materials
Junyi Cao
Evangelos Kalogerakis
Technical University of Crete
[arXiv]   [Code]   [Data]  


Abstract

We present SOPHY, a generative model for 3D physics-aware shape synthesis. Unlike existing 3D generative models that focus solely on static geometry or 4D models that produce physics-agnostic animations, our approach jointly synthesizes shape, texture, and material properties related to physics-grounded dynamics, making the generated objects ready for simulations and interactive, dynamic environments. To train our model, we introduce a dataset of 3D objects annotated with detailed physical material attributes, along with an annotation pipeline for efficient material annotation. Our method enables applications such as text-driven generation of interactive, physics-aware 3D objects and single-image reconstruction of physically plausible shapes. Furthermore, our experiments demonstrate that jointly modeling shape and material properties enhances the realism and fidelity of generated shapes, improving performance on generative geometry evaluation metrics.

Results
(Best viewed in a PC instead of a mobile device)
(Note: We display the simulation space in the animations as background axes to help visualize the effects of collisions.)
$\triangleright$ Image-to-4D Results $\triangleleft$
Input
                    View 1                    
                    View 2                    
Condition
DG4D [1]
Baseline
SOPHY
DG4D [1]
Baseline
SOPHY
image condition
DG4D Result
Unnoticeable motion.
image condition
Implausible geometry.
image condition
DG4D Result
Unnoticeable motion.
image condition
Implausible geometry.
image condition

Expected outcome: A chair is falling.

image condition
DG4D Result
Unnoticeable motion.
image condition
Incorrect material for the plant leaves (too stiff).
image condition
DG4D Result
Unnoticeable motion.
image condition
Incorrect material for the plant leaves (too stiff).
image condition

Expected outcome: Drag the leaves of a planter.

image condition
DG4D Result
Bad physical motion.
image condition
Incorrect material for the loveseat leg (too soft).
image condition
DG4D Result
Bad physical motion.
image condition
Incorrect material for the loveseat leg (too soft).
image condition

Expected outcome: Drop a love seat.

image condition
DG4D Result
Unnoticeable motion.
image condition
Incorrect material for the headband (too stiff).
image condition
DG4D Result
Unnoticeable motion.
image condition
Incorrect material for the headband (too stiff).
image condition

Expected outcome: Drop a headband.

$\triangleright$ Text-to-4D Results $\triangleleft$
                    View 1                    
                    View 2                    
STAG4D [2]
Baseline
SOPHY
STAG4D [2]
Baseline
SOPHY
DG4D Result
Unrealistic physical motion.
image condition
Incorrect material for the chair backrest (too soft).
image condition
DG4D Result
Unrealistic physical motion.
image condition
Incorrect material for the chair backrest (too soft).
image condition

Input Condition: A chair made of a polypropylene seat connector, a metal foot, a metal leg, a polyester backrest, and a polyester seat is dropping.

DG4D Result
Time-varied handbag size.
image condition
Misalignment with the condition "rigid handbag" (too soft).
image condition
DG4D Result
Time-varied handbag size.
image condition
Misalignment with the condition "rigid handbag" (too soft).
image condition

Input Condition: Throwing a rectangular, rigid handbag with a woven texture, dark brown leather-like material, metallic rivets, a central clasp, and two curved handles.

DG4D Result
Unnoticeable motion.
image condition
Misalignment with the condition "cotton bill" (too stiff).
image condition
DG4D Result
Unnoticeable motion.
image condition
Misalignment with the condition "cotton bill" (too stiff).
image condition

Input Condition: Throwing a hat made of a denim crown, and a cotton bill.

DG4D Result
Time-varied pillow size.
image condition
Misalignment with the condition "dark reddish-brown fabric" (blue).
image condition
DG4D Result
Time-varied pillow size.
image condition
Misalignment with the condition "dark reddish-brown fabric" (blue).
image condition

Input Condition: Throwing a pillow made of a dark reddish-brown fabric, and a dark gray-brown piping.

BibTeX
@article{Cao_2025_SOPHY,
    author    = {Cao, Junyi and Kalogerakis, Evangelos},
    title     = {{SOPHY}: Generating Simulation-Ready Objects with Physical Materials},
    journal   = {arXiv xxx},
    year      = {2025}
}
Reference

[1]: Ren et al. DreamGaussian4D: Generative 4D gaussian splatting. arXiv, 2023.

[2]: Zeng et al. STAG4D: Spatial-temporal anchored generative 4D gaussians ECCV, 2024.

This webpage is modified from the template provided here.