Skip to main content
Gizmo’s core capability is AI-driven asset generation: you describe an object in plain English and the system produces a fully articulated, physics-ready 3D asset in seconds. There’s no manual modelling, no rigging, and no need to hand-author collision geometry or inertia tensors — all of that comes out of the generation pipeline automatically.

​
How It Works

Asset generation follows a straightforward pipeline:
1

Write a text prompt

Describe the object you need. Include material, colour, size, and intended purpose for best results. For example: “Heavy-duty industrial steel pallet rack shelving unit, orange and blue”.
2

AI model processes your prompt

Gizmo’s generative model interprets the description and synthesises a 3D mesh with textures, materials, and articulation structure.
3

Physics properties are auto-generated

Collision geometry, mass, inertia tensors, joint limits, and friction parameters are automatically derived from the geometry — no manual configuration needed.
4

Asset is ready to use

The finished asset appears in your library. Add it to a scene, adjust its placement, or export it directly to Isaac Sim or MuJoCo.

​
What “Fully Articulated” Means

A fully articulated asset is one where every moveable or physically meaningful part of the object is properly represented in the simulation model:
  • Links — rigid bodies that make up the object’s structure (e.g., a stool’s seat, post, and base)
  • Joints — connections between links that define how they can move relative to each other (revolute, prismatic, fixed)
  • Collision geometry — simplified shapes that the physics engine uses for contact detection, distinct from the visual mesh
  • Physics properties — mass, centre of mass, inertia tensor, contact friction, and restitution coefficients
Fully articulated assets load directly into Isaac Sim or MuJoCo and behave physically from the first simulation step — no additional setup required.
This stands in contrast to “static mesh” assets, which have visual geometry but no physics properties and can’t be used in a dynamic simulation without significant manual work.

​
Writing Effective Prompts

The quality of the generated asset is directly shaped by the specificity of your prompt. Vague prompts produce generic results; detailed prompts produce assets that closely match your simulation requirements.
Include material, colour, size or scale, and intended use in every prompt. The more context you give the model, the closer the output will be to what you need.
Examples:
Instead of…Try…
pallet rackHeavy-duty industrial steel pallet rack shelving unit, orange and blue
stoolRolling medical stool with backrest, chrome base, and black vinyl seat
lampDainty brass pendant light with fluted glass shade
binRed step-on biohazard waste receptacle with warning label
bollardYellow steel safety bollard with black hazard stripes
Prompt checklist:
  • ✅ Specify the material (steel, wood, vinyl, brass, fabric…)
  • ✅ Include colour or finish (chrome, orange-and-blue, white marble…)
  • ✅ Mention function or context (medical, industrial, warehouse, domestic…)
  • ✅ Add distinguishing features (with backrest, with glass doors, with casters…)
  • ❌ Avoid single-word prompts — they produce overly generic geometry

​
Supported Asset Types

Gizmo’s generation model covers a broad range of object categories relevant to robotics simulation. Below are representative examples from each category.

Industrial & Warehouse

Objects found on factory floors and in fulfilment centres:
  • Pallet Rack — shelving for pallet storage
  • Safety Bollard — yellow steel post with hazard stripes
  • Rolling Ladder — steel warehouse ladder with casters and handrails
  • LED High Bay Light — round ceiling fixture with aluminium heat sink

Workstations & Furniture

Workbenches, seating, and storage for office or lab environments:
  • Packing Station — industrial workbench with built-in scale and tape dispenser
  • Kitchen Island — marble waterfall countertop with dark wood base
  • Nightstand — wooden nightstand with single drawer
  • Wall Cabinet — storage cabinet with glass doors and metal handles

Medical & Laboratory

Equipment for clinical and research simulation environments:
  • Rolling Stool — medical stool with backrest, chrome base, and vinyl seat
  • Biohazard Bin — red step-on receptacle with warning label
  • Medical Wall Cabinet — wall-mounted storage with glass doors

Lighting

Fixture types for realistic scene illumination and appearance:
  • Brass Pendant Light — fluted glass shade on a brass fitting
  • Bedside Lamp — small table lamp with fabric shade
  • LED High Bay — industrial round ceiling fixture

Retail & Commercial

Props for retail, food service, and public-space simulation:
  • Shopping Cart — standard metal wire cart with red plastic handle
  • Chalkboard Menu — framed board with hand-lettered text
Don’t see the category you need? Gizmo’s model handles a wide range of everyday and industrial objects. Try describing your asset — if generation quality is lower than expected, refine your prompt with more detail or reach out at viswajit@antimlabs.com with a feature request.
For a hands-on walkthrough of the generation flow, see the Generating Assets guide.