Description
BowerBot is an AI agent designed for the assembly of production-ready OpenUSD scenes from natural language instructions. Operating as a part of the production pipeline, it allows users to describe a scene in plain English, and it interprets spatial relationships, material choices and lighting moods to build the scene.
A notable feature of BowerBot is its compliance with Academy Software Foundation standards; every asset it incorporates is automatically structured in a studio-grade hierarchy.
The AI agent intelligently places six types of light - dome, sphere, rect, disk, cylinder, and distant lights. BowerBot operates with a built-in validation functionality, iterating and correcting errors to ensure scenes are structured correctly.
Specific commands allow scenes packaged for USDZ export to be used in platforms like Apple Vision Pro, NVIDIA Omniverse, or other USD-compatible viewers.
BowerBot extends its functionalities through a plugin system for asset providers, accepting contributions from the community and enabling connectivity to digital libraries such as Sketchfab.
It runs on OpenUSD 25.x, Python 3.12+, LiteLLM GPT-4.1, Claude Pydantic, and is available under Apache 2.0 license.
Key Features
ASWF-compliant USD Hierarchy
MaterialX Materials
Natural Language Scene Assembly
DCC Integrations
Pricing Plans
Releases
Reviews
Pros & Cons
Pros
Open-source
Builds OpenUSD scenes
Assembles from natural language
Cons
Requires NVIDIA Omniverse/Maya/Houdini/usdview/Apple Vision Pro
Requires understanding of natural language descriptions
Specific to VFX/animation/digital twins/spatial computing/AEC
