PhysX-Omni is a research framework for generating simulation-ready 3D assets that carry physical properties, not just geometry and texture. Most image-to-3D generators output a visually plausible mesh with no notion of joints, mass, or material behavior, so the result cannot be dropped into a physics engine or robotics simulator without manual rigging. PhysX-Omni targets three object classes in one pipeline: rigid bodies, deformable objects, and articulated objects with movable parts, producing outputs that convert directly to URDF-style scene descriptions for downstream simulation. The pipeline fine-tunes a Qwen2.5 vision-language model to reason about an object's physical structure from a single conditioning image, then decodes geometry with a TRELLIS-based decoder before a conversion step assembles the result into a simulation-ready scene file. Alongside the model, the project releases PhysXVerse, a dataset of physically annotated 3D assets, and PhysX-Bench, a benchmark with a VLM-based evaluation pipeline for scoring physical plausibility. The codebase builds on prior PhysX-Anything research and reuses components from TRELLIS, Qwen, and Grounded-Segment-Anything, distributed under a non-commercial research license.