USD Validation NVIDIA
September 10, 2026 · View on GitHub
An extensible framework to validate OpenUSD assets. Inspired by Pixar's usdchecker, this library extends validation capabilities with additional rules and provides automatic issue fixing.
It is the standalone successor to omniverse-asset-validator and is intended for use in any USD pipeline.
Features
- Rule-based validation — Modular rule interface with registration mechanism for custom validators
- Flexible engine — Run validation on a
Usd.Stage, individual layers, or recursively search folders - Auto-fixing — Issue fixing interface for applying automated corrections when rules provide suggestions
- Command line interface — Standalone CLI for validation outside of GUI applications
- Lightweight — Pure Python implementation requiring only OpenUSD
Installation
Install from PyPI:
pip install usd-validation-nvidia
Optional Dependencies
For full functionality, install with optional dependencies:
# Include usd-core
pip install usd-validation-nvidia[usd]
# Include USD Exchange
pip install usd-validation-nvidia[usdex]
# Include NumPy (for optimizations)
pip install usd-validation-nvidia[numpy]
# Install all optional dependencies
pip install usd-validation-nvidia[usd,numpy]
Repository Layout and Releases
This repository is the single source repository for the shared NSPECT release and publishes
two Python packages from the same commit, both versioned from the shared VERSION.md:
usd-validation-nvidia— the validation engine (source undervalidation/src/, with package docs, examples, and tests undervalidation/)usd-profiles-nvidia— profile/spec definitions and code generation tooling (profiles/directory, auvworkspace member; formerly a standalone repository)
uv build --all-packages produces wheels and sdists for both packages.
Build from Source
Generate the capabilities package with the in-repo profiles/ member (usd-profiles-nvidia), then build the wheel:
uv run \
--no-project \
--with ./profiles \
python -m usd_profiles_nvidia.codegen \
--docs-root specs \
--destination-dir validation/src \
--package-name usd_validation_nvidia.capabilities \
--reverse-domain com.nvidia.usd
uv build --all-packages -o dist
Basic usage
from usd_validation_nvidia import ValidationEngine, IssueFixer
engine = ValidationEngine()
asset = "path/to/asset.usda"
results = engine.validate(asset)
for issue in results.issues():
print(f"{issue.severity}: {issue.message}")
fixer = IssueFixer(asset)
fix_results = fixer.fix(results.issues())
for result in fix_results:
print(f"{result.status}: {result.issue.message}")
fixer.save()
The ValidationEngine also supports selecting specific rules with enable_rule() / disable_rule().
To filter by category in Python, enable the rules returned by CategoryRuleRegistry().get_rules(category).
Command Line Interface
The nvidia_usd_validate command provides validation from the terminal:
# Validate a single file
nvidia_usd_validate asset.usda
# Validate a directory recursively
nvidia_usd_validate ./assets/
# Apply automatic fixes
nvidia_usd_validate --fix asset.usda
# Validate specific categories
nvidia_usd_validate --category Material --category Geometry asset.usda
# Enable specific rules only
nvidia_usd_validate --no-init-rules --rule StageMetadataChecker asset.usda
# Export results to CSV
nvidia_usd_validate --csv-output results.csv asset.usda
# Show help
nvidia_usd_validate --help
Documentation
Requirements
- Python 3.10 - 3.13 with OpenUSD 23.11 or later
- Python 3.10 - 3.13 with USD Exchange 3.0.0 or later
- Python 3.14 with OpenUSD 26.08
License
Apache-2.0 AND CC-BY-4.0
AI Coding Agents
The AGENTS.md file and .agents/skills directory contain structured guidance for AI coding agents. Start there for Python project setup, local venv setup, and generated requirement validation workflows.