Skip to main content
Everything you need to know about Gizmo — what it is, who it’s for, what you can build with it, and how to get the most out of it during the alpha. If your question isn’t answered here, reach out directly and the team will help.

​
General

Gizmo is an AI-powered 3D scene authoring tool built by AntiM Labs. It lets you generate fully articulated 3D assets for robotics simulators from plain text descriptions — no manual modeling required.You describe what you want (e.g., “a height-adjustable packing station with conveyor belt and safety railing”), and Gizmo’s AI generates a 3D asset complete with joints, links, collision geometry, and physics properties, ready to drop into your simulation environment.Gizmo is currently in alpha and under active development.
Gizmo is designed for people who build and work with robotics simulations:
  • Robotics engineers who need simulation-ready environment assets without spending hours in a 3D modeling tool
  • Simulation developers building virtual environments in NVIDIA Isaac Sim or MuJoCo
  • Researchers who need to rapidly prototype diverse scene configurations for training, testing, or benchmarking robot policies
  • Anyone who wants to go from a text description to a physics-ready 3D asset as fast as possible
If you spend time manually building or sourcing assets for robot simulation, Gizmo is built for you.

​
Simulators & Formats

Gizmo currently supports two industry-standard robotics simulators:
  • NVIDIA Isaac Sim — export as USD or URDF
  • MuJoCo — export as MJCF/XML
Support for additional simulators is on the roadmap. If there’s a specific simulator you’d like to see supported, let the team know at viswajit@antimlabs.com.
Gizmo supports the following export formats:
SimulatorFormatNotes
NVIDIA Isaac SimUSDRecommended; preserves full scene hierarchy and materials
NVIDIA Isaac SimURDFUse for ROS-compatible Isaac Sim pipelines
MuJoCoMJCF / XMLNative MuJoCo model description format
Select your target format in the export dialog before downloading.
A fully articulated asset is one that is simulation-ready out of the box — not just a static mesh. When Gizmo generates an asset, it automatically produces:
  • Joints — the degrees of freedom between moving parts (e.g., hinges, sliders, revolute joints)
  • Links — the rigid body segments that make up the asset
  • Collision geometry — simplified shapes used by the physics engine for contact detection
  • Physics properties — mass, inertia, friction, and damping values
This means you can drop the exported asset into Isaac Sim or MuJoCo and immediately simulate it with realistic physics, without needing to manually rig or configure anything.

​
Access & Pricing

Gizmo is currently in a gated alpha. Access is granted on a request basis.To request access, send an email to viswajit@antimlabs.com with:
  • Subject: Gizmo Alpha Access Request
  • A brief description of your use case or what you’d like to build
The team will review your request and follow up to grant access.
Yes. Early API access is available on request for developers who want to integrate Gizmo’s asset generation into their own pipelines or tooling.To request API access, email viswajit@antimlabs.com with:
  • Subject: Gizmo API Access Request
  • A brief description of how you intend to use the API
The API is in early stages, and early adopters will have direct input into its design.
Pricing and access terms during the alpha are handled on a case-by-case basis. For the most up-to-date information on access and pricing, contact the team directly at viswajit@antimlabs.com.

​
Using Gizmo

The more specific, the better. Gizmo’s AI uses every detail in your prompt to shape the generated asset. Vague prompts produce generic or incomplete results; detailed prompts produce accurate, usable assets.When writing a prompt, try to include:
  • Material — e.g., “stainless steel”, “black ABS plastic”, “powder-coated aluminum”
  • Color — e.g., “bright yellow”, “matte gray”, “natural wood finish”
  • Dimensions — e.g., “1.8 meters tall”, “60cm wide”, “standard countertop height”
  • Purpose or context — e.g., “used in a warehouse”, “hospital-grade”, “retail display fixture”
  • Key features — e.g., “with four swivel casters”, “adjustable shelves”, “hinged door on front”
Weak prompt: a binStrong prompt: a heavy-duty biohazard waste bin, bright yellow, 90cm tall, rigid plastic, with foot-pedal lid opener and a labeled warning symbol on the front
Gizmo is designed for the kinds of real-world objects found in industrial, commercial, and research environments — the kinds of things robots actually interact with.Examples of assets you can generate:
  • Warehouse packing stations and conveyor fixtures
  • Rolling ladders and mobile work platforms
  • Biohazard bins and medical waste containers
  • Kitchen islands and food service equipment
  • Retail display fixtures and shelving units
  • Industrial lighting rigs and overhead fixtures
  • Medical devices and IV stands
  • Laboratory equipment and workbenches
If you’re unsure whether a particular asset type is supported, try a prompt and see — or reach out to the team for guidance.
Your feedback during the alpha directly shapes what Gizmo becomes. Whether it’s a bug, a feature idea, a prompt that didn’t work as expected, or a simulator compatibility issue — the team wants to hear it. Email viswajit@antimlabs.com anytime.